of re P uo se R re tn I ht e M oc e D ya le R - Aaltodoc

147
g n i k r o w t e N d n a s n o i t a c i n u m m o C f o t n e m t r a p e D f o s i s y l a n A e c n a m r o f r e P d n a n o i t a c o l l A e c r u o s e R n o i t a g i t i M e c n e r e f r e t n I d n a b n I r o f s d o h t e M d r a w r o F d n a e d o c e D g n i y a l e R h a l l U m a n I L A R O T C O D S N O I T A T R E S S I D

Transcript of of re P uo se R re tn I ht e M oc e D ya le R - Aaltodoc

-otl

aA

DD

5

21/

810

2

+gggai

a*GMFTSH

9 NBSI 6-6608-06-259-879 )detnirp( NBSI 3-7608-06-259-879 )fdp( NSSI 4394-9971 )detnirp( NSSI 2494-9971 )fdp(

ytisrevinU otlaA

gnireenignE lacirtcelE fo loohcS gnikrowteN dna snoitacinummoC fo tnemtrapeD

if.otlaa.www

+ SSENISUB YMONOCE

+ TRA

+ NGISED ERUTCETIHCRA

+ ECNEICS

YGOLONHCET

REVOSSORC

LAROTCOD SNOITATRESSID

hall

U ma

nIdn

a ed

oce

D dn

abnI

rof

sdoh

teM

noit

agit

iM

ecne

refr

etnI

dna

noi

taco

llA

ecru

ose

R fo

sisy

lan

A ec

namr

ofre

P

gni

yale

R dr

awr

oF y

tisr

evi

nU

otla

A

8102

gnikrowteN dna snoitacinummoC fo tnemtrapeD

fo sisylanA ecnamrofreP dna noitacollA ecruoseR

noitagitiM ecnerefretnI dnabnI rof sdohteM drawroF dna edoceD

gniyaleR

hallU manI

LAROTCOD SNOITATRESSID

-otl

aA

DD

5

21/

810

2

+gggai

a*GMFTSH

9 NBSI 6-6608-06-259-879 )detnirp( NBSI 3-7608-06-259-879 )fdp( NSSI 4394-9971 )detnirp( NSSI 2494-9971 )fdp(

ytisrevinU otlaA

gnireenignE lacirtcelE fo loohcS gnikrowteN dna snoitacinummoC fo tnemtrapeD

if.otlaa.www

+ SSENISUB YMONOCE

+ TRA

+ NGISED ERUTCETIHCRA

+ ECNEICS

YGOLONHCET

REVOSSORC

LAROTCOD SNOITATRESSID

hall

U ma

nIdn

a ed

oce

D dn

abnI

rof

sdoh

teM

noit

agit

iM

ecne

refr

etnI

dna

noi

taco

llA

ecru

ose

R fo

sisy

lan

A ec

namr

ofre

P

gni

yale

R dr

awr

oF y

tisr

evi

nU

otla

A

8102

gnikrowteN dna snoitacinummoC fo tnemtrapeD

fo sisylanA ecnamrofreP dna noitacollA ecruoseR

noitagitiM ecnerefretnI dnabnI rof sdohteM drawroF dna edoceD

gniyaleR

hallU manI

LAROTCOD SNOITATRESSID

-otl

aA

DD

5

21/

810

2

+gggai

a*GMFTSH

9 NBSI 6-6608-06-259-879 )detnirp( NBSI 3-7608-06-259-879 )fdp( NSSI 4394-9971 )detnirp( NSSI 2494-9971 )fdp(

ytisrevinU otlaA

gnireenignE lacirtcelE fo loohcS gnikrowteN dna snoitacinummoC fo tnemtrapeD

if.otlaa.www

+ SSENISUB YMONOCE

+ TRA

+ NGISED ERUTCETIHCRA

+ ECNEICS

YGOLONHCET

REVOSSORC

LAROTCOD SNOITATRESSID

hall

U ma

nIdn

a ed

oce

D dn

abnI

rof

sdoh

teM

noit

agit

iM

ecne

refr

etnI

dna

noi

taco

llA

ecru

ose

R fo

sisy

lan

A ec

namr

ofre

P

gni

yale

R dr

awr

oF y

tisr

evi

nU

otla

A

8102

gnikrowteN dna snoitacinummoC fo tnemtrapeD

fo sisylanA ecnamrofreP dna noitacollA ecruoseR

noitagitiM ecnerefretnI dnabnI rof sdohteM drawroF dna edoceD

gniyaleR

hallU manI

LAROTCOD SNOITATRESSID

seires noitacilbup ytisrevinU otlaASNOITATRESSID LAROTCOD 521 / 8102

ecruoseR fo sisylanA ecnamrofreP noitagitiM ecnerefretnI dna noitacollA

dna edoceD dnabnI rof sdohteM gniyaleR drawroF

hallU manI

fo rotcoD fo eerged eht rof detelpmoc noitatressid larotcod A eht fo noissimrep eht htiw ,dednefed eb ot )ygolonhceT( ecneicS

cilbup a ta ,gnireenignE lacirtcelE fo loohcS ytisrevinU otlaA 72 no loohcs eht fo SAUT 1UT llah erutcel eht ta dleh noitanimaxe

.)noon ta( kcolc'o 21 ta 8102 enuJ

ytisrevinU otlaA gnireenignE lacirtcelE fo loohcS

gnikrowteN dna snoitacinummoC fo tnemtrapeD

rosseforp gnisivrepuS neniälämäH iryJ rosseforP

rosivda sisehT

awgnufatuM drawdE .rD

srenimaxe yranimilerP neniälämäH omiT rosseforP daassA damahoM rosseforP

tnenoppO

itarsumlE demmahoM rosseforP

seires noitacilbup ytisrevinU otlaASNOITATRESSID LAROTCOD 521 / 8102

© 8102 hallU manI

NBSI 6-6608-06-259-879 )detnirp( NBSI 3-7608-06-259-879 )fdp( NSSI 4394-9971 )detnirp( NSSI 2494-9971 )fdp(

:NBSI:NRU/if.nru//:ptth 3-7608-06-259-879

yO aifarginU iknisleH 8102

dnalniF

tcartsbA otlaA 67000-IF ,00011 xoB .O.P ,ytisrevinU otlaA if.otlaa.www

rohtuA hallU manI

noitatressid larotcod eht fo emaNdnabnI rof sdohteM noitagitiM ecnerefretnI dna noitacollA ecruoseR fo sisylanA ecnamrofreP

gniyaleR drawroF dna edoceD

rehsilbuP gnireenignE lacirtcelE fo loohcS

tinU gnikrowteN dna snoitacinummoC fo tnemtrapeD

seireS seires noitacilbup ytisrevinU otlaA SNOITATRESSID LAROTCOD 521 / 8102

hcraeser fo dleiF gnireenignE snoitacinummoC

dettimbus tpircsunaM 8102 hcraM 6 ecnefed eht fo etaD 8102 enuJ 72

)etad( detnarg hsilbup ot noissimreP 8102 enuJ 7 egaugnaL hsilgnE

hpargonoM noitatressid elcitrA noitatressid yassE

tcartsbAelibom G5 dna G4 eht rof noisnetxe gnisimorp a sedivorp tpecnoc )CER( ralulleC decnahnE yaleR

lacol htiw sedon gniyaler )FD( drawroF-dna-edoceD ,drager siht nI .smetsys noitacinummoc na yalp dna ecnamrofrep krowten devorpmi elbane rewop noissimsnart wol dna egarevoc

fo sisylana ecnamrofrep eht no sesucof siseht sihT .smetsys elibom erutuf eht ni elor tnatropmi ew ,krowemarf lacitcarp a sA .krowten ralullec-orcam gniyalrevo eht rednu gniyaler FD poh-laud 01 esaeleR )ETL( noitulovE mreT gnoL )PPG3( tcejorP pihsrentraP noitareneG drihT eht redisnoc

.gniyaler FD xelpud-flah dnabni 1 epyT dnoyeb dna

kniL sseccA eht dna )LR( kniL yaleR eht neewteb )AR( noitacollA ecruoseR redisnoc ew ,tsriF )EU( tnempiuqE resU eht neewteb noitcennoc )e2e( dne-ot-dne eht rehtegot esirpmoc taht )LA( AR tnereffid fo yduts evitarapmoc a tuo yrrac eW .)ETL ni BNe sa dellac osla( noitats esab eht dna

ecuded ew ,noitaulave ecnamrofrep eht roF .AR lamitpo ecruoser ni sucof niam htiw ,semehcs .snoitalumis hguorht stluser yfirev dna ,setar naem dna egatuo e2e eht rof snoisserpxe mrof-desolc

ecnerefretni secruoser oidar ecracs ot euD .LR ni noitagitim ecnerefretni eht yduts ew ,dnoceS

ezylana dna ecudortni ew ,eroferehT .etar atad e2e eht rof rotcaf gnitimil a semoceb ylisae LR ni eht etagitim ot desu eb nac taht noitamrofni edis detimil htiw euqinhcet gnimrofmaeb elpmis a ni yllaicepse taht wohs ,salumrof citylana no desab ,stluser ecnamrofreP .LR eht ni ecnerefretni

edis tib wef ylno htiw detucexe ylevitceffe eb nac noitagitim ecnerefretni eht lennahc naiciR evorpmi ecnerefretni LR fo noitagitim ehT .noitats esab gnirefretni eht ot yaler morf noitamrofni

.soiranecs ynam ni etar atad e2e eht ylraelc

tnemevorpmi egarevoc gnidliub-ni eht rof yaler cidamon roodtuo fo esu eht no yduts esac a ,drihT ssenrah yam yticapac dna egarevoc gnidliub-ni roop eht ycnegreme .g.e fo esac nI .tuo deirrac si yaler cidamon gniyolpme yb dexaler eb nac melborp sihT .srednopser ycnegreme fo snoitarepo eht tpecnoc CER hcus fo ecnamrofrep eht redisnoc eW .ycnegreme na fo etis eht ot esolc thguorb si taht

elbaton a edivorp nac yaler cidamon taht dnuof si tI .sepyt gnidliub tnereffid eerht fo esac ni .sresu gnidliub-ni eht rof tnemevorpmi ecnamrofrep

ecnahne ot tpecnoc evitceffe na sedivorp gniyaler taht wohs seiduts denoitnem eht morf stluseR

laicurc fo si ti weiv fo tniop ecnamrofrep e2e morf ,teY .yticapac dna egarevoc smetsys elibom ecnerefretni LR eht no dna ,LA dna LR neewteb AR eht no tneps si erac laiceps taht ecnatropmi

sdrowyeK ,noitagitiM ecnerefretnI ,MRR ,luahkcab yaleR ,gniyaleR drawroF-dna-edoceD dnabnI egarevoC roodnI

)detnirp( NBSI 6-6608-06-259-879 )fdp( NBSI 3-7608-06-259-879

)detnirp( NSSI 4394-9971 )fdp( NSSI 2494-9971

rehsilbup fo noitacoL iknisleH gnitnirp fo noitacoL iknisleH raeY 8102

segaP 241 nru :NBSI:NRU/fi.nru//:ptth 3-7608-06-259-879

To my Mother, Father and Family Members

Preface

This Doctoral thesis work has carried out in the Department of Communica-

tion and Networking at Aalto University School of Electrical Engineering, Es-

poo, Finland. The Academy of Finland, Ericsson, and Cassidian have funded

this research work.

First, I would like to highly appreciate and express my sincere and utmost

gratitude to my supervisor, Prof. Jyri Hämäläinen for his all-time generous

unceasing support and continous encouragements throughout my research

activities in his group. I would like to mention that Prof. Jyri Hämäläinen is

the one who open the door of research world for me and due to his honest

guidance; I was able to achieve this big milestone of my life.

I would like to say my special thanks to my thesis instructor, Dr. Edward

Mutafungwa for his all-time available support and help during my doctoral

research work, during which we had done several informative discussions as

well as help me in writing/publication of all my research manuscripts. His

expertise enable an opportunity to polish my technical writing and presenta-

tion skills to present research contributions.

I am thankful to Dr. Zhong Zheng for his utmost support and patience who

provide me a continous guidance in solving problems at all phases of my re-

search work in master and doctoral studies.

I am thankful to Dr. Alexis Dowhuszko for his cooperative approach and

very informative feedback on various research issues throughout my research

studies, even in his busy schedules.

I would like to say my special gratitude to Dr. David González González

for his politeness, humbleness and being continously helpful at every phase

of my research work.

I extend my thanks to Udesh Oruthota, Haile Beneyam, Konstantinos Koufos,

Aamir Mahmood, and Umar Saeed for their tips and suggestions throughout

my research phases.

i

Preface

I would like to acknowledge and highly appreciate the efforts and help of the

support team during my whole work duration in Aalto University. I am grate-

ful and thankful to Viktor Nässi for his all-time any-time welcoming smile and

cooperation. I am really thankful to Marja Leppäharju, Kati Voutilainen, Mari

Paloheimo, Liukko Heli, Haaranen Sirpa, Rinne Essi, Patana Sanna, Kiveliö

Sari, Hietala Juhapekka, Laaksonen Joni and Lehtola Timi for their utmost

and all time support during my stay in Aalto University.

Espoo, Finland, June 7, 2018,

Inam Ullah

ii

Contents

Preface i

Contents iii

List of Abbreviations and Symbols v

1. Introduction 1

1.1 Motivation and problem definition . . . . . . . . . . . . . . . . 1

1.2 Scope of the Thesis . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.3 Summary of Thesis Contributions and Publications . . . . . . 4

1.4 Author’s Contributions . . . . . . . . . . . . . . . . . . . . . . . 5

1.5 Structure of the Thesis . . . . . . . . . . . . . . . . . . . . . . . 6

2. Relaying in LTE-Advanced Systems and Beyond 7

2.1 3GPP LTE, LTE-Advanced and LTE-A Pro Systems . . . . . . . 9

2.2 Relaying Principles and Classifications . . . . . . . . . . . . . . 15

2.3 LTE-A Relaying System . . . . . . . . . . . . . . . . . . . . . . . 21

2.4 Relaying beyond LTE-A . . . . . . . . . . . . . . . . . . . . . . 25

3. Resource Optimal Relaying 31

3.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

3.2 Previous work and Contributions . . . . . . . . . . . . . . . . . 31

3.3 System Model and Resource Allocation Schemes . . . . . . . . 34

3.4 Performance analysis . . . . . . . . . . . . . . . . . . . . . . . . 40

3.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

4. Practical Interference Mitigation for the Relay Backhaul Link 57

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

4.2 Previous Work and Contributions . . . . . . . . . . . . . . . . . 58

4.3 System Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

iii

Contents

4.4 Analysis of SINR and e2e Outage Rate . . . . . . . . . . . . . . 66

4.5 Performance Evaluation . . . . . . . . . . . . . . . . . . . . . . 72

4.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78

5. Rapidly Deployable Relays for Indoor Environments 81

5.1 Background and Motivation . . . . . . . . . . . . . . . . . . . . 81

5.2 Previous work and Contributions . . . . . . . . . . . . . . . . . 82

5.3 Description of the relay deployment cases . . . . . . . . . . . . 85

5.4 System Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87

5.5 Performance Evaluation and Simulations . . . . . . . . . . . . 92

5.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99

6. Conclusions and future work 101

6.1 Analysis of Optimal Resource Sharing . . . . . . . . . . . . . . 101

6.2 Interference Mitigation for the Relay Backhaul Link . . . . . . 102

6.3 Rapidly Deployable Relays for Outdoor-to-indoor Coverage . 103

References 105

iv

List of Abbreviations and Symbols

Abbreviations

1G First Generation

2G Second Generation

3G Third Generation

3D Three Dimensional

3GPP Third Generation Partnership Project

4G Fourth Generation

5G Fifth Generation

Ant Antenna

AF Amplify and Forward

AL Access Link

AP Access Point

AAS Active Antenna Systems

AS Antenna Selection

BER Bit Error Ratio

BF Beamforming

BS Base Station

BW Bandwidth

CA Carrier Aggregation

CAGR Compounded Annual Growth Rate

CDF Cumulative Distribution Function

CN Core Network

COST COopération européenne dans le domaine de la recherche Sci-

entifique et Technique

CoMP Coordinated Multipoint Transmission and Reception

CP Cyclic Prefix

CQI Channel Quality Information

v

List of Abbreviations and Symbols

CSI Channel State Information

D2D Device-to-Device

DAS Distributed Antenna Systems

DC Dual Connectivity

DF Decode and Forward

DeNB Donor Evolved Node-B

DLBF Dual-layer Beamforming

DL Downlink

DRX Discontinuous Reception

dB decibel

EBF Elevation Beamforming

EDGE Enhanced Data rates for GSM Evolution

EE Energy Efficiency

EH Energy Harvest

e2e End-to-End

eIMTA Enhanced International Mobile Telecommunication Ad-

vanced

eMBMS Evolved Multimedia broadcast Multicast Service

eNB Evolved Node-B

FBMC Filter Bank Multicarrier

FDD Frequency Division Duplex

FD Full Duplex

FD-MIMO Full-dimension MIMO

GPRS General Packet Radio Service

GSM Global System for Mobile Telephony

Gbps Giga bits per second

HD Half Duplex

Hetnet Heterogeneous Networks

HSPA High Speed Packet Access

Hz Hertz

ICIC Inter-Cell Interference Coordination

ICPWF Iterative Co-phasing Waterfilling

IL Implementation Loss

IM Interference Mitigation

IMT-2000 International Mobile Telecommunications-2000

IMT-A International Mobile Telecommunications-Advanced

IoT Internet of Things

IP Internet Protocol

vi

List of Abbreviations and Symbols

IRT Intelligent Ray Tracing

ISD Inter-Site Distance

ISI Inter-Symbol Interference

ITS Intelligent Transportation System

ITU-R International Telecommunication Union-Radio

kHz kilo Hertz

L1 Layer 1

L2 Layer 2

L3 Layer 3

LA Link Adaptation

LAA License Assisted Access

LoS Line-of-Sight

LPN Low Power Nodes

LTE-A Long Term Evolution-Advanced

LTE Long Term Evolution

LTE-A Pro Long Term Evolution-Advanced Pro

LWA LTE-WLAN Aggregation

MAC Media Access Control

MBS Macro Base Station

MBMS Multimedia Broadcast Multicast Services

MBSFN Multimedia Broadcast over Single Frequency Network

MHz Mega Hertz

MIMO Multiple-Input Multiple-Output

MME Mobility Management Entity

MMF Max-Min Fairness

MNO Mobile Network Operator

MRN Moving Relaying Node

MTC Massive Machine-Type Communication

MUE Macro User Equipment

MUST Multi-User Superposition Transmission

Mbps Megabits per second

mmW Millimeter Wave

msec Millisecond

NB-IoT Narrow-band Internet-of-Things

NLoS Non-Line-of-Sight

NOMA Non-Orthogonal Multiple Access

OFDM Orthogonal Frequency Division Modulation

OFDMA Orthogonal Frequency Division Multiple Access

vii

List of Abbreviations and Symbols

QCP Quantized Co-phasing

PAPR Peak to Average Power Ratio

PDF Probability Distribution Function

PHY Physical layer

PL Path Loss

PLMN Public Land Mobile Networks

PLR Packet Loss Ratio

PRB Physical Resource Block

ProSe Proximity Service

RAN Radio Access Network

REC Relay Enhanced Cellular

RF Radio Frequency

RL Relay Link

RN Relay Node

RRH Remote Radio Head

RRM Radio Resource Management

RR Round Robin

RS Receiver Selection

RUE Relay User Equipment

Rx Receive

S-GW Serving Gateway

SNR Signal-to-Noise-Ratio

SINR Signal-to-Interference-and-Noise-Ratio

SON Self Organizing Networks

TDD Time Division Duplex

TDMA Time Division Multiple Access

TP Throughput

TTI Transmit Time Interval

Tx Transmit

UDN Ultra Dense Networks

UE User Equipment

UL Uplink

UMTS Universal Mobile Telecommunications System

V2I Vehicle-to-Infrastructure

V2N Vehicle-to-Network

V2P Vehicle-to-Pedestrian

V2V Vehicle-to-Vehicle

V2X Vehicle-to-any-thing

viii

List of Abbreviations and Symbols

WG Working Groups

WI Walfisch Ikegami

WLAN Wireless Local Area Networks

ix

List of Abbreviations and Symbols

Symbols

2σ2 Power of the fading component

A(·) Antenna transmission pattern

Am Antenna’s front-to-back ratio

a Indoor propagation loss (dB/m)

Beff Bandwidth efficiency

BPRB Bandwidth of a PRB

Cav Average capacity

D Perpendicular distance between external antenna and external

wall

Da Amount of data for all UEs on AL

Da,u Amount of data for uth UE on AL

Dmin Minimum data rate requirement

Dr Amount of data on RL

Ei(·) Exponential integral function

E(·) Expectation

F (·) Cumulative distribution function

f(·) Probability distribution function

f1 Frequency 1

f2 Frequency 2

fc Carrier frequency

h(1)q Channel coefficient at best receive antenna for a qth transmit

antenna of eNB

h(2)q Channel coefficient at worst receive antenna for a qth transmit

antenna of eNB

hjlq Channel coefficient at the lth receive antenna of the RN from

the qth transmit antenna of the jth eNB

Irestn,k Rest of interfering eNBs signals

Kd Rician K factor for desired DeNB

Ki Rician K factor for interfering eNB

k RN of the nth cell

L Number of hops

Nd Number of phase bits in the dedicated link

Ni Number of phase bits in the interfering link

Na Number of PRBs allocated for AL

NASF Number of subframe allocated for AL in 10 msec frame

Nmax Maximum number of PRBs

x

List of Abbreviations and Symbols

NMBSFN Number of MBSFN subframes in a 10 msec frame

NPRB Number of PRBs in a subframe

NPRB,u Numbers of PRBs allocated to uth UE

Pn Noise power

P out Outage probability

PRx,DeNB Power received by RN from DeNB

PRx,Other eNBs Power received by RN from neighboring eNBs

Pt Transmission power

Q3 3rd Quarter

Re2e e2e rate

Ra Instantaneous rates on AL

Rmin Minimum rate

Rout Outage rate

Routa Outage rate on AL

Rr Instantaneous rates on RL

r1 Target relay node served by DeNB

r2 Relay node served by interfering eNB

S Distance between transmitter and receiver

Sa,u Spectral efficiency of AL

Sr Spectral efficiency of RL

T Total duration of TTI

Ta AL portion of TTI

Tr RL portion of TTI

τa Relative transmission time allocated for AL

τr Relative transmission time allocated for RL

τmaxr Relative transmission time threshold for RL

U Number of UEs

Wa AL bandwidth

Wr RL bandwidth

W Precoding complex weight vector

wjq Precoding weight applied at the qth transmit antenna of the

jth eNB

γa Instantaneous AL SNR

γa Mean AL SNR

γd Mean power from dedicated signal on RL from desired DeNB

γi Mean power of interfering signal from interfering eNB

γr Mean RL SNR

γn,n,k Mean power of the desired DeNB on RL

xi

List of Abbreviations and Symbols

γm,n,k Mean power of the interfering eNB on RL

γn,n,k Mean power of the dominant interfering signal on RL

ν2 Power of the static signal component

θ Angle of direction of main beam of antenna from UE towards

transmitter

θ3dB Angle of antenna beam which is 3dB lower than the main

beam of directional antenna

ΥB RL SINR for baseline scheme

ΥBF RL SINR for beamforming schemes

Υa,u SINR level for uth UE on AL

Υeff SINR efficiency

Υr SINR level on RL

xii

1. Introduction

1.1 Motivation and problem definition

1.1.1 Development towards small cell systems

Mobile communication has become an integral part of our everyday life. Dur-

ing the last two decades drivers for the evolution, has been the increasing

number of mobile broadband subscribers and their growing demand for larger

capacities due to increasing number of smart-phones and other data con-

sumption devices [1]. For example, there was almost 7.6 billion mobile sub-

scriptions globally at first quarter (Q1) of 2017, there are 2.1 billion Long Term

Evolution (LTE) subscriptions and the number is expected to reach 5 billion by

year 2022 [2].

To that end, the standardization organizations such as Third Generation Part-

nership Project (3GPP) and International Telecommunication Union Radio commu-

nication (ITU-R) are continuously specifying new technological solutions to

meet the future services demands. These technologies include the widely-

deployed fourth generation (4G) and evolved 4G technology enhancements

that support services ranging from mobile broadband connectivity (with peak

throughput up to 1 Gbps) to human-to-machine and machine-to-machine-

type communication. Moreover, it is envisioned that the future fifth gener-

ation (5G) technologies will contribute towards a Networked Society, whereby,

all kind of services will be provided through wireless connectivity [1, 3]. Ac-

cordingly, it is expected that the number of 5G subscription will exceed half

a billion by 2022 [2]. To that enhanced use cases enabled by extended mo-

bile broadband connectivity (eMBB), massive Internet-of-Things (IoT) (also

referred to massive machine-type communications, mMTC) and ultra-reliable

low-latency communications (URLLC), will cater to the exploding demands

1

Introduction

for capacity along with other growing range of vertical applications and busi-

ness models [4].

The service, capacity and ubiquitous coverage targets of future wireless sys-

tems can be in part fulfilled by increasing the available spectrum and im-

proving the spectral efficiency. Yet, even more substantial growth of net-

work capacity can be achieved by reusing spectrum through network den-

sification [5, 6]. The commonly considered network densification is obtained

through the heterogeneous network (HetNet) deployment where Low Power

Nodes (LPNs) with less than 10 W transmission power complement the tradi-

tional high-power (20 W or higher transmission power) macro-cell sites [6,7].

Term LPN is used here to draw a distinction between the complementary

small nodes and the legacy macro-cell sites. A vast range of LPNs exists de-

pending on the use scenario attributes (area size, indoor or outdoor location,

etc.). The LPNs include small nodes such as relays, pico/micro nodes, remote

radio heads and distributed antenna systems. It is estimated by Cisco that 60%

of the global cellular traffic will be offloaded to various types of LPNs in near

future [8].

1.1.2 Wireless LPN: Relaying

For LPN network extensions the connectivity to the core network is a crucial

problem and a number of LPN backhauling solutions has been identified de-

pending on the deployment scenario. For indoor LPNs there is the possibility

to leverage in-building wireline infrastructure such as fiber and twisted cop-

per pairs (digital subscriber lines), to provide backhaul. For outdoor LPNs sit-

uation is more complicated since access to backhaul lines might not be possi-

ble without new cable installations. Therefore wireless backhauling solutions

such as fixed microwave or millimeter wave radio links and even satellite links

have been considered as backhaul options for the outdoor LPNs. Yet, these

backhauling alternatives are usually far too expensive when compared to the

LPN traffic volumes.

To that end, the LTE-Advanced (LTE-A) relaying specification provides an

interesting backhauling scenario [9]. The proposed LTE-A relays are self-

backhauled towards the serving Donor evolved Node B (DeNB) via a stan-

dardized radio (Un) interface. The relay deployment scenarios considered in

3GPP standardization discussion include the indoor coverage enhancement,

outdoor coverage extension and temporarily deployed (nomadic) relays. An

important and technically challenging class of LTE-A relays are the inband

dual-hop Decode-and-Forward (DF) Relay Nodes (RNs). The self-backhauling

2

Introduction

ability of dual-hop DF inband relaying protocol provides an opportunity to

efficiently utilize existing radio resources by exploiting the same frequency

band for both Relay Link (RL) between DeNB and RN, and Access Link (AL)

between RN and UE.

However, self-backhauling through inband radio resources may cause a ca-

pacity bottleneck on the RL which obviously affects the End-to-End (e2e) link

performance between UE and DeNB. To compensate the scarce radio resources

in RL efficient Resource Allocation (RA) schemes are needed to optimally

share the radio resources between the RL and AL. Moreover, the inband na-

ture of relaying in frequency reuse 1 systems like 4G may lead to serious co-

channel interference challenges. The interference on RL is crucial since it will

impact to the service of all users connected to the RN. If RNs are fixed part of

the network infrastructure, as is the case in 4G, then the interference in RL is

not temporary in nature but may occur almost continuously. Fortunately, the

interference between the transmitter and receiver with fixed locations can be

effectively mitigated through e.g. beamforming techniques.

The self-backhauling property of relays also provides some new application

options. In e.g. an emergency event, unexpected and high traffic demands

may occur locally in the network. Especially indoor coverage and capacity

may be unacceptable low when the emergency responders and public author-

ities in place need it most. For such situations nomadic RN provides a flexible

solution that can be used to increase local capacity. Actually, a nomadic RN

can be installed to a vehicle and switched on in the place of emergency. The

outdoor location on top of a vehicle (or even on top of a telescope mast) sup-

port good RL while AL indoor coverage can be provided by placing the RN

close to the building wall.

1.2 Scope of the Thesis

The specific focus of the thesis is on the dual-hop inband decode-and-forward

relaying. This type of relaying is part of 4G LTE standards and currently con-

sidered for 5G as well. While mathematical analysis plays an important role in

the performance evaluation, all analytical results have been verified through

simulations. Although most of the results are quite generic in nature, they

are practically viable and can be used in the development of future relaying

systems for 5G. We also note that underlying assumption is that relays are

overlaid by a macro-cellular network.

As discussed briefly in the previous section, the main challenges of the in-

3

Introduction

band relaying are the scarcity of radio resources and the co-channel interfer-

ence. These challenges have been main drivers of the thesis. Accordingly, the-

sis considers the RA between the RL and AL for inband relaying with aim to

maximize the end-to-end (e2e) performance. Since results from the resource

allocation study show that the RL easily becomes the e2e performance bottle-

neck, the practical interference mitigation in the RL has been the second main

subject for the thesis. Finally, a case study is provided to investigate a rapidly

deployable outdoor RN to improve the local in-building coverage and capac-

ity in case of e.g. emergency events. In addition, in this case the co-channel

interference and resource allocation for relaying plays an important role.

1.3 Summary of Thesis Contributions and Publications

The main contributions of this thesis are briefly described below.

• Contribution 1

The RA between RL and AL can be carried out by many different ap-

proaches. In literature fixed RA between RL and AL has been widely de-

ployed while studies on resource optimal RA are rare. In this approach in-

stantaneous RA between RL and AL is deployed that maximizes the e2e

throughput. In this thesis we present a comparative analysis of RA schemes

including the fixed RA with and without buffer, and the resource optimal

RA. For that purpose, we deduce closed-form expressions for the mean and

outage e2e rate. In case of resource optimal RA the mean rate attains an

expression in terms of an integral that does not admit a closed-form solu-

tion. Therefore, we derive a tight lower bound that accurately approximates

the mean e2e data rate performance. In addition, we deduce closed-form ex-

pressions of e2e rate for the case where DL and UL communication between

source and destination is decoupled. Main results have been summarized

in [10] and [11].

• Contribution 2

As discussed, the co-channel interference in RL can become crucial since

resources in RL are scarce. Accordingly, we study the interference mitiga-

tion in RL based on Channel State Information (CSI) fed back from RN to

the both serving and interfering base stations. The main emphasis is on

the analytical investigation but again, all results have been verified through

simulations. We derive analytical expressions for the RL SINR distribution

4

Introduction

assuming the Rice and Rayleigh fading combinations on the RL and the in-

terfering links. Based on the results also e2e outage rate is analytically for-

mulated. These contributions have been summarized in [12] while the work

in [13] served as a starting point for the study.

• Contribution 3

Rapidly deployable RN is an interesting concept where a nomadic out-

door RN can be located e.g. within the proximity of a building with an

emergency and used to provide an in-building coverage for emergency re-

sponders. To study the impact of RN location, overlay network interference

and RA in such scenario three different indoor environments has been con-

sidered. Namely we assume a 3GPP 5×5 grid, 3GPP dual strip model and

a more realistic deployment case based on ray tracing propagation model-

ing. Performance is evaluated in terms of e2e throughput. Results have been

previously presented in [14].

1.4 Author’s Contributions

The author of this thesis was in the leading role in the research reported in

publications [10], [13], [14] and in the submitted manuscript [12]. Dr. Alexis

Dowhuszko has been leading the research reported in the manuscript [11].

In [10], [12], [13] and [14] author of this thesis carried out performance simu-

lations, participated actively in the mathematical analysis by deducing part of

the formulas and was the main author while writing the manuscript. Dr. Ed-

ward Mutafungwa (in [12], [13], [14]), Dr. David Gonzalez G.(in [10], [13]), Dr.

Alexis Dowhuszko (in [10]), Dr. Zhong Zheng (in [10], [14]) and Dr. Beneyam

Haile (in [13]) have given support in the mathematical analysis, writing and

in the proof-reading phase while supervising professor Jyri Hämäläinen has

been helping to identify problems and to model the system. He has been also

guiding the publication work.

The author of this thesis also actively contributed to the manuscript [11],

where his focus was especially in the simulations and paper writing.

5

Introduction

1.5 Structure of the Thesis

Chapter 2 provides a brief overview of 3GPP LTE system covering the prin-

ciples of the basic physical layer features like orthogonal frequency division

multiple access (OFDMA), radio resource management and inter-cell interfer-

ence coordination. Discussion also includes LTE-A radio features and basics

of the relaying therein. To that end, relaying modes and protocols are be-

ing explained in a general level while particular focus is placed on the LTE-A

Type 1 inband relaying scheme. Furthermore, resource scheduling mecha-

nism in relaying and benefits of relaying are briefly discussed.

Chapter 3 presents a comparative study of different RA schemes employed

to distribute the radio resource between RL and AL. Main focus is on the per-

formance analysis of the resource optimal RA that is compared to fixed RA

with and without buffer. The rate distributions and mean rate formulas are

presented to support the analysis of the relaying e2e data rate performance.

In addition, this chapter also contains the derivation of the closed-form ex-

pression for the e2e data rate in case where the DL and UL transmissions are

decoupled. Material presented in this chapter is based on the publication [10]

and the manuscript [11].

Chapter 4 investigates the mitigation of the RL interference through simple

transmit beamforming techniques. Chapter is started by a literature survey

of the field and followed by the problem modeling and analysis that provides

derivation of analytical formulas for the SINR and the outage probability in

the RL by assuming the Rice and Rayleigh fading combinations for the RL

and the interfering links. Analytical study is complemented by simulations to

verify the analytical formulas and provide further insights to the system per-

formance. Material presented in this chapter is based on the publication [12]

while the work in [13] complements the study.

Chapter 5 presents a performance evaluation for a nomadic RN concept

from the perspective of outdoor-to-indoor coverage in different building sce-

narios. The simulation campaign provides the e2e throughput performance

experienced by indoor UE in different building and RN deployment situa-

tions. Material presented in this chapter is based on the publication [14].

Finally, Chapter 6 summarizes the contributions of this thesis.

6

2. Relaying in LTE-Advanced Systemsand Beyond

This chapter focuses on the relaying systems specified in LTE-A standards,

as well as, on the development of relaying technologies thereof. First, we

introduce the baseline LTE system with special emphasis on relaying. This

background discussion includes a thorough introduction of the 3GPP LTE

including the architecture and specifications emerging from the ITU-R ini-

tiative International Mobile Telecommunication-2000 (IMT-2000). Although, the

LTE technology is already deployed, its evolution continues as can be seen

from LTE-A and its future releases (e.g., Release 14) [1,15]. While current LTE

systems support various mobile applications, the customer expectations for

high data rates with low latency are constantly increasing. Accordingly, both

ITU and 3GPP are working on upcoming fifth generation (5G) technologies

while evolving the current LTE releases (to Release 14) in parallel with de-

velopment of new radio access technologies. The LTE Release 14 and beyond

are required to support the upcoming 5G requirements [15–17]. Figure 2.1

shows a schematic diagram of the evolution phases occurred in the mobile

communication technologies including the LTE standard releases.

GSM

1990 2000 2008 2011 2015

GPRSEDGE

UMTSHSPA

HSPA+

LTELTE-A

Voice-centric 1G/2G Mobile broadband 3G/4G

5G

~2020

Networked society

LTE-A Pro

Figure 2.1. Evolution time-line of the mobile wireless communication technologies

7

Relaying in LTE-Advanced Systems and Beyond

Relaying

Relay Backhaul Enhance-

ments Issues

• Network Planning & Op-

timization

� [18–24]

• RRM Schemes

� [20, 22, 25–29]

• MIMO, CoMP & Interfer-

ence Mitigation

� [12, 13, 30–35]

• Path Loss & Antenna

Elevation

� [36–39]

Mobile System Enhance-

ments

• Coverage Extension

� [18, 23, 27, 33, 40–51]

• Capacity Enhancements

� [13, 14, 19, 21, 24, 27, 33, 42,

43, 46, 48, 49, 52–58]

• Energy Efficiency

� [59–66]

• Performance comparison

Studies

� [13, 14, 43, 44, 46, 48, 50, 52,

54, 61, 67–71]

Future Research

• Open Issues & Future

Challenges

� [5, 13, 35, 38, 72–84]

PHY/MAC Layers

• Channel Waveforms

� [76, 82, 83, 85–92]

• Massive MIMO

� [92–100]

• Relaying with Network

Coding

� [101–109]

• Cooperative Communica-

tion & Relay Selection

� [78, 110–120]

• Channel Access

� [117, 118, 120–123]

Classification of Relays

based on

• Resource Allocation

� [1, 124, 125]

• Duplexing Modes

� [126–128]

• Relaying Processing

� [9, 21, 45, 47, 52, 68, 129–

138]

• Deployment Modes

� [3, 139–141]

Deployment & Usage sce-

narios

• UDN & Hetnets

� [3, 6, 17, 71, 142]

• D2D Communication

� [1, 80, 81, 143–147]

• M2M/IoT

� [3, 148–150]

• Moving Networks

� [140, 141, 147, 151, 151–

157]

• Rapid Deployments

� [14, 81, 144, 145, 155, 158–

160]

Figure 2.2. Overall Structure of our literature survey of Relaying systems.

Figure 2.2 presents the overall structure of our literature survey on relaying

in LTE-A systems and beyond. We have used the mentioned references in the

forthcoming discussion. Yet, due to practical reasons we are not following in

the following sections exactly the same categorization as in Figure 2.2.

The rest of the chapter is organized as follows. Section 2.1 gives the overall

summary of LTE releases including the system requirements, architecture as

well as highlights enhanced physical layer technologies introduced in each re-

lease. Section 2.2 explains the relaying principle and classification of several

8

Relaying in LTE-Advanced Systems and Beyond

relaying modes employed in wireless communication. Furthermore, the em-

ployment of relaying in LTE-A system are being briefly described in section 2.3

with special focus on inband relaying. Moreover, this section also highlights

the relaying performance gains obtained in terms of coverage extension and

capacity enhancements. Finally, section 2.4 presents the overall literature sur-

vey on relaying beyond LTE-A systems including the open issues and future

challenges.

2.1 3GPP LTE, LTE-Advanced and LTE-A Pro Systems

2.1.1 Baseline LTE System

System Requirements and Architecture

The LTE standards were driven by the need for a long term efficient network

coverage and capacity solution [161]. The LTE standards initially targeted sys-

tem improvements over the preceding High-Speed Packet Access (HSPA) Re-

lease 6 standards. These initial targets include an improved instantaneous

peak data rates of 50 Mbps and 100 Mbps in uplink (UL) and downlink (DL),

respectively, a flexible operating bandwidths from 1.4 MHz up to 20 MHz and

less than 10 milliseconds (msec) round-trip delay over the LTE air interface.

Other targets include improvements in battery efficiency for LTE UE, reduced

cost per transmitted bit, support for high-speed mobility and compatibility

with legacy network architectures. LTE systems may operate on both paired

and unpaired spectrum, i.e. Frequency Division Duplexing (FDD) and Time

Division Duplexing (TDD). Moreover, the system can be operated on differ-

ent frequency bandwidths: 1.25 MHz, 1.6 MHz, 2.5 MHz, 5 MHz, 10 MHz,

15 MHz, 20 MHz.

Unlike the preceding 3GPP architecture, LTE systems are designed to en-

able a seamless Internet Protocol (IP) based connectivity between User Equip-

ment (UE) and Core Network (CN). This leads to a reduced number of net-

work interfaces. For example, radio base station (called as eNB in 3GPP termi-

nology) is the only intermediate node between the UE and CN. It is noted that

LTE eNB uses Uu to serve its UEs. This flat architecture decreases the amount

of signaling and jitters. The main components of the LTE system architec-

ture are shown in Figure 2.3. As already mentioned, the LTE Radio Access

Network (RAN) enables a flat system architecture where eNB is responsible

for all the radio related functionalities of the network. We note that eNB em-

9

Relaying in LTE-Advanced Systems and Beyond

ploys the so-called X2 interface to communicate with its neighbouring eNBs,

in order to enable a seamless active mode mobility in the network [1].

eNB1

eNB3

eNB2UE

UE

UEX2

X2 X2

S1 S1

S1

S1

MME/S-GWMME/S-GW

Internet/PLMN

EPC

E-UTRAN

SGi SGi

Figure 2.3. LTE network architecture.

LTE Radio Interface

The LTE system applies a multi-carrier transmission scheme known as Or-

thogonal Frequency-Division Multiplexing (OFDM) in order to enable broad-

band services. In OFDM, data symbols are modulated over orthogonal

narrow-band subcarriers with subcarrier spacing of 15kHz. This minimizes

the selectivity of wide band channels which may cause the loss of data sym-

bols. Moreover, OFDM scheme also minimizes the Inter-Symbol Interfer-

ence (ISI) through the use of Cyclic Prefix (CP) technique, enabling a simple

receiver design .

The temporal resources are segmented into radio frames comprising of 10 sub-

frames with duration of 10 msec. Each sub-frame is formed by two slots of

duration of 0.5 msec. Each slot consists of 6 or 7 OFDM symbols. The Physical

Resource Block (PRB) is the smallest element used in the transmission resource

allocation with bandwidth of 180 kHz. Each PRB contains 12 consecutive sub-

carriers for one 0.5 msec slot. One or more contiguous resource blocks can be

allocated to a physical channel.

Multi-antenna transmission techniques hold a key role in LTE technology.

To that end, even the first LTE eNBs and UEs employ dual antennas in or-

der to improve the downlink performance via Multiple Input Multiple Out-

put (MIMO) methods. Multiple antennas can be used to increase data rates

10

Relaying in LTE-Advanced Systems and Beyond

and/or suppress the interference [1].

LTE systems also exploit the rapid channel variations through dynamic

channel-dependent scheduling. Here, the time-frequency resources are dy-

namically assigned to UEs per Transmission Time Interval (TTI). Due to multi-

path fading, the UE radio link may experience rapid instantaneous variations

in both time and frequency domains [1]. This channel condition information

is known as Channel-State Information (CSI) and it is estimated using the refer-

ence signals in the DL direction. Once CSI is conveyed to eNB it can be applied

in the scheduler while scheduling the resources to UEs.

To increase the spectrum efficiency, LTE employs the frequency reuse-one.

Thus, all the neighbouring transmission points can use the same spectrum.

In DL this leads to a notable interference between the eNBs. To that end, co-

ordinated transmission between eNBs has been introduced. This technique

is known as Inter-Cell Interference Coordination (ICIC). Hence, eNBs exploit the

X2 interface to exchange the control messages with neighbouring eNBs to co-

ordinate the overlapping transmissions [1].

2.1.2 LTE-Advanced System

The baseline LTE system was defined in standard Releases 8 and 9. Though,

there was a need to further develop the system in order to meet the rapidly in-

creasing service demands. To that end, LTE-Advanced (Release 10) represents

the evolution of LTE that is designed to achieve IMT-Advanced requirements.

LTE-A system aims to achieve peak data rates of up to 1 Gbps (for low mobil-

ity users) and 500 Mbps in DL and UL, respectively. The target peak spectral

efficiencies are 30 bps/Hz and 15bps/Hz in DL and UL, respectively. LTE-A

enhances the cell edge user throughput (5th%-ile user throughput) in order to

achieve a more homogeneous user experience in the cell [162].

In addition to these requirements, 3GPP also demanded the backward-

compatibility with LTE systems and seamless support for LTE Release 8 UEs.

To that end, LTE-A standards introduced a number of enhancements which

are depicted in Figure 2.4 and described briefly thereafter [1, 124, 163].

11

Relaying in LTE-Advanced Systems and Beyond

LTE-

Adv

ance

d

Release – 10/11

Downlink (8x8) / Uplink (4x4) MIMO Coordinated Multipoint (CoMP)Heterogeneous networks (Hetnet)Carrier Aggregation (CA)Relaying

Release – 12

Enhanced Downlink MIMOSmall Cell ON/OFFDual ConnectivityMachine Type Communication (MTC)Proximity Service (ProSe) or Device-to-Device (D2D) CommunicationSelf Optimizing Networks (SON)Hetnet MobilityMultimedia Broadcast/Multicast ServiceEnhanced International Mobile Telecommunication Advanced (eIMTA)Frequency Division Duplex – Time Division Duplex Carrier Aggregation (FDD-TDD CA)

Figure 2.4. LTE-A communication technologies.

Carrier Aggregation

LTE release 10 introduces a new bandwidth extension technique known as

Carrier Aggregation (CA), where multiple frequency components, each with

different bandwidths are aggregated. CA enables a wider bandwidth for

transmission purposes both in DL and UL. In principle this allows the ag-

gregation of up to five component carriers and transmission at maximum on

100 MHz aggregated band [1,124]. The number of component carriers in CA

is extended in later releases up to 32.

Multi-Antenna Enhancements

In LTE releases, various multi-antenna techniques have been introduced. To

that end, Release 10 introduced the enhanced DL MIMO and UL MIMO

schemes. The DL MIMO transmission may apply up to 8 transmission layers

while in UL spatial multiplexing of up to 4 transmission layers can be used [1].

Coordinated Multi-Point

LTE Release 11 introduced the so-called Coordinated Multipoint Transmis-

sion (CoMP). In CoMP, different transmission points (in same or different cell

sites) coordinate their transmission and reception to create less interference

for their users and to enhance the performance. In CoMP, network may uti-

lizes dynamic point selection such that the transmission can be switched be-

tween involved transmission points. Alternatively network can utilize the

joint transmission/reception, where the transmitted/received data signals

12

Relaying in LTE-Advanced Systems and Beyond

are jointly processed to enhance the transmission/reception performance [1].

Heterogeneous Deployments

In Heterogeneous Network (HetNet), the mixture of access nodes with

different transmission powers are deployed under the macro-overlaid net-

work (usually include high power macro node with transmission antenna el-

evation above the rooftop level). In HetNet low-power nodes (usually pico

nodes) are overlaid by macro-cell coverage area with purpose to match the

service provision with the non-uniform demand. In LTE Release 10 new ways

were introduced to coordinate the interference among different layers of Het-

Net with different transmission powers.

Furthermore, the Hetnet concept considered in Release 12 enable a dense de-

ployment of small cells to achieve high network throughput. In dense de-

ployment, there is a high likelihood that device is in the service area of many

small cells. Such overlapping downlink transmissions certainly create heavy

interference. Accordingly, LTE Release 12 introduced a technique of turning

ON/OFF the small cells, so as to minimize the inter-cell interference as well

as to improve the energy efficiency [1].

Machine Type Communication

Machine Type Communication (MTC) applies wireless connectivity for ma-

chine communication under the control of an access node. MTC can be cate-

gorized into two major groups namely, to massive MTC and to critical MTC.

The former type comprise less critical, less complex and low power consump-

tion wireless devices, such as, sensor, actuator, etc. These type of devices have

long battery life even with span of 10 years [3]. The latter type represents a sta-

ble, always-on and reliable wireless connection. Related devices are used in

mission critical scenarios such as traffic control, industrial applications, smart

grid and so on [1, 3].

Device-to-Device Communication

In LTE Release 12, a new non network-centric communication mechanism

was introduced known as Device-to-Device (D2D) communication. In D2D,

the communication devices establish a direct communication link with each

other rather then connect via cellular base station. To that end, two usage

scenarios are public safety communication and local commercial usage. In

the former case, the D2D development includes the in-coverage and out-of-

coverage communication while in the latter case, each device is required to

discover its neighbouring devices [1].

13

Relaying in LTE-Advanced Systems and Beyond

Relaying

Relaying deploys a low power node known as Relay Node (RN) which acts

as an intermediate node between user and network elements. Specifically,

RN utilizes wireless self-backhaul link towards the eNB known as a Donor

eNB (DeNB). The RN deployment reduces the UE-Infrastructure distance and

eventually, minimizes the experienced path loss of signals between UE-DeNB.

In the following, the backhaul link between RN and DeNB is named as Relay

Link (RL) which carries both the UE data traffic as well as the control signaling

for RNs. The RL have normal LTE air interface characteristics [164]. Similarly,

the Direct Link and Access Link (AL) refer to the DeNB-UE link and the RN-UE

link, respectively. Figure 2.5 shows the schematic diagram of REC network.

Section 2.3 presents in more detail the LTE-A relaying system.

DeNB

RN

UE

Figure 2.5. LTE-A relay enhanced cellular network.

2.1.3 3GPP LTE-Advanced Pro

The LTE system evolution has been further extended by two releases namely,

by Release 13 (closed in March, 2016) and Release 14, also known as LTE-A

Pro. The newly introduced concepts are summarized in Figure 2.6.

In Release 13, the MTC performance was further improved by introducing

enhancement techniques. For example, the MTC enabled devices can be op-

erated at frequency spectrum of 1.4 MHz with transmission power 20 dBm

and battery life upto 10 years [3]. This not only reduce the device cost but

also improves the energy efficiency. Another LTE-based track is known as

Narrow-band Internet-of-Things (NB-IoT) further evolving the LTE-A systems

by exploiting the 180 kHz carrier. Moreover, the 3GPP communication tech-

nologies are envisioned to play a significant role in the deployment of Intel-

ligent Transportation System (ITS) by enabling the wireless connectivity be-

tween vehicles and any thing around it, in order, e.g., to ensure the public

safety of transportation system. To that end, the LTE enabled Vehicle-to-any-

thing (LTE-V2X) deployment scenarios. This X-any-thing could either be a

14

Relaying in LTE-Advanced Systems and Beyond

vehicle (V2V), pedestrian (V2P), network (V2N) or Infrastructure (V2I) across

the road side [15, 165].

LTE-

A P

ro

Release – 13Active Antenna Systems (AAS)Elevation Beamforming (EBF) and FD-MIMOEnhanced Signaling for CoMPEnhanced D2D/ProSeEnhancement for MTC & Narrow-band Internet-of-Things (NB-IoT)Indoor Positioning EnhancementEnhancement in CA/License-assisted Access (LAA)DL Multi-User Superposition Transmission (MUST)RAN sharing EnhancementDual Connectivity (DC)LTE-WLAN Aggregation (LWA)RAN enhancements for extended Discontinuous Reception (DRX) in LTE

Release – 14Enhanced FD-MIMO with 32 Tx AntennasLAA in Uplink TransmissionReduction in Latency timeIntelligent Transportation System via LTE-V2X: Vehicle-to-Vehicle (V2V), Vehicle-to-Infra (V2I), & Vehicle-to-Pedestrian (V2P) Enhanced operation in Unlicensed SpectrumEnhanced Positioning/Network assistance aspectsEvolved Multimedia broadcast Multicast Service (eMBMS)Superposition Coding for Enhanced DL Multiuser Transmission

Figure 2.6. LTE-A Pro communication technologies.

In LTE Release 13, the efficient utilization of available spectrum is further de-

veloped by employing the CA over the licensed and unlicensed frequency

spectrum (usually 5 GHz license free frequency is used). This approach is

known as License-Assisted Access (LAA) [1,124]. Furthermore, Release 13 in-

troduced the Dual Connectivity (DC) framework that enables a wireless de-

vice to simultaneously communicate with two eNBs. This provides an oppor-

tunity to aggregate the user-planes and a wireless device can communicate

with the multiple eNBs both in UL and DL in order to obtain improved per-

formance. Moreover, this also add robustness to the network when device is

controlled by one eNB while transmitting the user data to other eNBs. In DC,

UE can select the best eNB for UL and DL communication [1].

2.2 Relaying Principles and Classifications

Many types of RNs have been proposed since the earliest RN were first intro-

duced [3, 166]. RNs can be classified according to various criteria, such as,

15

Relaying in LTE-Advanced Systems and Beyond

(a) Classification based on Resource Allocation

(b) Duplexing-Based Classification

(c) Classification based on Relay Processing

(d) Classification based on Deployment Modes

In this section, we recall briefly the RN technologies under each category.

2.2.1 Classification based on Resource Allocation Strategies

Relaying can be classified according to how the RN resources are being shared

between RL and AL. In relay communication, four different unidirectional

communication links are required to enable communication between UE and

eNB both in UL and DL as shown in Figure 2.5. Different relaying protocols

are explained below.

Inband Relaying

In inband relaying, RL and AL utilize the same frequency spectrum, though,

links are separated via time division multiplexing on subframe basis as shown

in Figure 2.12. This type of relaying is spectral efficient due to usage of the

same frequency channel for both RL and AL operations. To minimize the in-

terference between the two links, RL and AL should be properly separated in

a time domain. This time division multiplexing of both links certainly affect

the system capacity as well as add some delay because simultaneous trans-

missions on RL and AL are not possible [1, 124, 125].

Outband Relaying

In outband relaying, RL and AL utilize separate frequency bands as shown in

Figure 2.12. In this type of relaying, RL and AL transmissions are not affect-

ing each other due to sufficient separation in frequency domain. This type of

relaying improves the system capacity due to availability of large spectrum

for each individual link but at the cost of inefficient use of frequency spec-

trum [1,124,125].

2.2.2 Duplexing-based Classification

Relaying can apply one of the given two duplexing modes:

16

Relaying in LTE-Advanced Systems and Beyond

Half-Duplex Relaying

In half duplex (HD) relaying, RN enables the transmissions (to DeNB in UL /

to UE in DL) and reception (from UE in UL / from DeNB in DL) on the same

frequency at different time instants. This requires suitable resource schedul-

ing in order to efficiently distribute the RN resources for UL and DL trans-

missions and receptions. In HD relaying, there is no co-channel interference

between RL and AL due to use of orthogonal time slots. Schematic diagram

of half-duplex relaying is shown by Figure 2.7.

DeNB

RN

UE DeNB

RN

UE

First Phase Second Phase

Figure 2.7. Half-duplex relaying mode.

Full-Duplex Relaying

In full duplex relaying (FDR), RN is able to carry out transmissions (to DeNB

in UL / to UE in DL) and reception (from UE in UL / from DeNB in DL), si-

multaneously either on the different or even the same frequency channel. The

former FDR scheme is known as out-band FDR [167] and the latter relaying

scheme is known as inband FDR [168], see Figure 2.8.

DeNB

RN

UE DeNB

RN

UE

Outband FDR Inband FDR

f 1f 1f 1f 2

Figure 2.8. Outband and Inband full-duplex relaying.

If RN exploit the same frequency spectrum for RL and AL transmissions

and receptions, then HD RN may suffer from the self-interference. That

is, the RN transmission interferes the RN receiving antenna However, this

self-interference can be mitigated by enabling enough spatial isolation be-

tween RN transmit and receive antennas. In case of out-band FDR, this

self-interference can be avoided by using large separation among the fre-

quency bands used for RN transmission and receptions [126]. The inband

17

Relaying in LTE-Advanced Systems and Beyond

FDR is required to employ the advanced interference mitigation schemes

as well as to isolate the transmit/receive antennas for combating the self-

interference [127,128].

2.2.3 Classification based on Relay Processing

Amplify-and-forward Relaying

Amplify-and-Forward (AF) RN, also known as repeater, amplify the analog

signals received from the source and retransmit it towards the destination as

shown in Figure 2.9 (a). During the AF relaying, AF also do amplify the in-

terference and noise in addition to the desired signal. This deteriorates the

overall system SINR level and limits the system throughput. Hence, these

RNs are more beneficial in the high SNR region. AF RNs are transparent to

both the source and destination and that is why they can be easily deployed in

the existing networks [68,129]. AF relaying can be categorized into two types

in terms of relay gain i.e., Variable Gain (VG) and Fixed Gain (FG) RNs. The

former protocol adjust its amplification factor according to the instantaneous

channel state while the latter protocol keeps a constant amplification factor

that is set by exploiting the statistical channel state information [130–132]. The

AF RN can also be considered as layer 1 RN [47, 134]. In WiMAX standard-

ization, this type of relaying are being proposed with the name of transparent

relaying [136].

Decode-and-forward Relaying

In DF relaying, RN first receives the entire signal from the source and then

decode and retransmit it to the destination as shown in Figure 2.9 (b). In DF

relaying, RN removes the interference and noise from the received signal af-

ter the first hop. Thus, this type of RN enables a good performance also in

low SNR environments [1]. Due to decoding and retransmitting of signals, it

causes time delay and increases the system complexity. In dual-hop DF relay-

ing system, throughput can be maximized on RL and AL, if both links enables

equal amount of data rate [21, 45, 133]. The DF RN can also be considered as

layer 2 RN [47]. Moreover, Layer 3 type RN, in addition to the Layer 3 func-

tionalities, comprises all base station functionalities like L2 RN. In WiMAX

this type of relaying is known as non-transparent relaying [136].

18

Relaying in LTE-Advanced Systems and Beyond

DeNB RN

(a) AF Relaying

UE

RF Signal Received

RF SignalRe-transmitted

Received SignalAmplified

DeNB RN

(b) DF Relaying

UE

RF Signal Received

RF SignalRe-transmitted

Received SignalDecodedEncodedAmplified

Figure 2.9. (a) Amplify-and-forward relaying and (b) Decode-and-forward relaying

Compress-and-forward Relaying

In Compress-and-Forward (CF) relaying (also known as estimate-and-

forward or quantize-and-forward), the RN operations fall in between the

AF and DF processing. Unlike the AF and DF relaying, where RN just re-

transmits the copy of what it receive from the source, the CF RN retransmits

the quantized form of the received message towards the destination. Hence,

the destination node uses the quantized message as well as the message re-

ceived from the source for decoding the source data. RN is not necessarily

required to decode the message received from the source but rather it needs

to extract the information which could be useful for the decoding purposes in

destination. Comparatively, CF relaying may outperform DF relaying when

the RL experience poor quality and vice versa [137,138].

2.2.4 Classification Based on the Deployment Mode

RNs can also be differentiated from the movement ability perspective. Mobil-

ity provides an additional characteristic to RN for enabling need-based cov-

erage in a target geographic location. Below, we give three different relaying

modes namely, fixed (or infrastructure), nomadic and mobile relaying [139].

Fixed Relaying

Fixed RN are usually permanently deployed at a fixed geographic location

with a purpose of providing coverage in a given (small) area. Conventionally

this type of RNs are deployed to enhance the service provision by improving

the UE SINR experienced at macro cell coverage holes and urban locations

19

Relaying in LTE-Advanced Systems and Beyond

affected by the shadowing. Fixed RNs can be also deployed at the macro

cell edge to systematically extend the network coverage/capacity. RNs can be

mounted in the lampposts or building walls or roof tops. Usually, the RN an-

tenna heights are lower than DeNB antenna heights in order to avoid mutual

interference between AL transmissions. In the network planning, RN deploy-

ment flexibility can be exploited to achieve a LOS conditions for RL [3, 139].

Schematic diagram of fixed relaying is shown by Figure 2.10.

DeNB

RN

UE

Coverage extension

Coverage holeShadowing

Urbanenvironment

Figure 2.10. Fixed (infrastructure) relaying scenarios.

Nomadic Relaying

Nomadic relaying is characterized by a semi-static RN deployment. RN can be

temporarily deployed in a certain location, where the cellular coverage is poor

and there is a temporary need for high capacity. These RNs are designed so

that they can be transported to the desired location and start operations with-

out complex configuration efforts. For example, nomadic relaying is attractive

solution in emergency/disaster situations where the emergency responders

and authorities are unable to (locally) communicate due to the unavailability

or congestion in the network. The UE served by a nomadic RN may experi-

ence both LoS and NLoS wireless channel condition on the AL. The nomadic

RN should be light and power efficient [3, 139].

Mobile Relaying

In mobile relaying, the RN is mounted on a moving vehicle (e.g., Train, Bus,

Tram), to enable on-board cellular services with good quality. A mobile RN is

connected with DeNB via mobile link, while it may have a static AL connec-

tivity to UEs located inside the vehicle. The mobile RL will comparatively add

more complexity in systems due to its entry and exit in/out of cell coverage

areas. The antenna height of the mobile RN would be relatively low because

of vehicle restrictions and other operational safety scenarios. Further details

20

Relaying in LTE-Advanced Systems and Beyond

on mobile relaying can be found from [3,139].

2.3 LTE-A Relaying System

2.3.1 Architecture of Relay Enhanced Cellular Network

Relaying is an add-on type of amendment in LTE, introduced first time in

LTE-A Release 10. To ensure the backward compatibility with earlier LTE re-

leases, RN acts like a normal eNB from the UE perspective. Similarly, core net-

work considers RN as an additional sector in DeNB coverage domain which

makes RN transparent for the CN as well. From the neighbouring eNB per-

spective, RN is also transparent and it assumes that UE is served by DeNB.

Moreover, in UL the RN utilizes the DeNB to forward the UE data towards

CN and neighbouring eNBs via standard interfaces (i.e., via S1 and X2) [124].

More information about the 3GPP relaying can be found from 3GPP standard-

ization technical reports [9, 100, 169–173].

eNBDeNB

UEUE

X2

S1S1

MME/S-GW

Internet/PLMN

REC Network

RNUE

X2 S1

Figure 2.11. LTE-A relay enhanced cellular network architecture.

In 3GPP LTE-A standards, relaying is differentiated as Type 1 and Type 2 re-

laying due to its capability of operation within the LTE-A network [9].

Type 1 Relay

In Type 1 relaying, RN is capable of controlling its own cell and all its RRM

functionalities. RN utilizes the local UE reporting to enable resource schedul-

ing among its own UEs. It sends its own physical cell identity with all con-

trol/synchronization channels and reference signals to UE. Type 1 RN acts as

21

Relaying in LTE-Advanced Systems and Beyond

a normal eNB to all type of UEs including the LTE Release 8 UEs. However,

RN is required to connect to CN via DeNB. Moreover, UE also treat RN as

normal eNB and send all the channel related information to RN. The L3 RN

can be considered as Type 1 RN. This type can be further sub-divided into

Type 1a and Type 1b RNs. The difference between Type 1a and Type 1b is

that Type 1a RN utilizes the outband relaying characteristics enabling RL and

AL transmissions on different frequency carriers. While the Type 1b utilizes

the inband relaying characteristics provided that RN transmit and receive an-

tennas, are sufficiently isolated [9]. Figure 2.12 presents the different LTE-A

relaying types.

DeNBRN

Out-band Relaying Type 1a

DeNBRN

Inband Relaying Type 1

DeNBRN

Inband Relaying Type 1b

Relay Backhaul Antenna

Figure 2.12. LTE-A relaying types.

Type 2 Relay

Type 2 RN utilizes the inband relaying characteristics for its operation in the

cellular network. This RN type operates in a transparent mode within the

existing donor cell, meaning that UEs are unaware of Type 2 RN in the cell.

That is, RN doesn’t create its own serving cell and does not have the physical

cell identity. These RNs are deployed to improve the SINR level and through-

put within the donor cell. Examples are the smart repeaters, DF RN and L2

RN [9]. The L2 relaying is not supported in LTE standards.

2.3.2 Radio Frame Structure in LTE relaying

This section describes the resource allocation when inband Type 1 RN is de-

ployed. Resource allocation scheme is required to share the available radio

resources among Direct link, RL and AL. The RL competes for the resources

with UEs served directly by DeNB on the direct link. Moreover, the RL and

AL utilize the same frequency bandwidth in inband relaying. To that end,

a resource allocation is required to isolate the transmission of both links, for

example by multiplexing the transmission in time domain. This multiplexing

22

Relaying in LTE-Advanced Systems and Beyond

should ensure that RN can carry out transmission on AL at a time, when the

RL transmission is on hold and the other way around. Figure 2.13 shows the

schematic diagram of the LTE downlink radio frame for the inband relaying

case.

As it can be seen in Figure 2.13, DeNB allocates (in this example case) three

subframes for RL among the ten subframes in a radio frame. The RN is hold-

ing its transmission on AL during these three subframes to avoid the self-

interference. In UL, it is possible for RN to ignore the transmission coming

from UEs on AL while communicating with DeNB.

Since LTE-A RN is backward compatible with the Release 8, it sends a cell-

specific reference and control signals in all DL subframes by utilizing the so-

called MBSFN capability of LTE technology [174]. More precisely, RN trans-

mits all cell-related information in the first one or two OFDM symbols of a

MBSFN subframe in DL. Accordingly, UEs are not in MBSFN mode and thus

they ignore MBSFN subframes where RL is carried out. By this, RN can quit

the transmission towards UEs and start receiving transmission from DeNB.

The remaining seven subframes in Figure 2.13 are mutually shared by the

DeNB and RN to schedule their own corresponding UEs on direct link and

AL, respectively. This however, creates interference among the overlapping

eNB and RN layers [1, 124, 175].

MUE

DeNB

RN

RUE

0 1 3 4 5 6 7 8 92

DeNB transmission to MUEs

DeNB backhaul transmission to RN on MBSFN subframe

RN reception of DeNB backhaul transmission on MBSFN subframe

RN access link transmission to RUEs

Blank subframes due to MBSFN DeNB interference to RUEs

Figure 2.13. Radio resource splitting among the direct link, RL and AL at DeNB

It is noted that the RN also receives co-channel interference from the sur-

rounding eNBs when they are serving their own RNs on the same RL sub-

frames.

23

Relaying in LTE-Advanced Systems and Beyond

2.3.3 Benefits of LTE-A Relaying

Relaying provides benefits in terms of coverage extension and capacity en-

hancement as compared to legacy cellular networks. Below Table 2.1 indicates

several studies on advantages of relaying. Moreover, this section also summa-

rizes briefly the literature regarding the studies done on various enhancement

techniques.

Table 2.1. References describing the LTE-A relaying benefits

References Scenarios Methodology Benefits

[27, 42, 44,

46, 51, 52,

61, 68, 69]

Macro and/or

Pico Only & REC

networks

SimulationsCoverage extension &

capacity gains

[13, 14, 19,

21, 23, 47,

48, 56, 58,

62, 71, 176]

Macro Only &

REC networks

Analytical

and/or

Simulations

Capacity

enhancements

[43, 45, 49,

54, 67, 177]

Macro Only &

REC networksSimulations Coverage extension

[50, 70, 71]Macro Only &

REC networksSimulations

Interference mitigation

& Throughput gains

[37, 40, 41]Macro Only &

REC networks

Field Tests &

measurements

Coverage extension in

indoor/coverage holes

[57, 59–66]

Macro Only

and/or REC

networks

Analytical

and/or

Simulations

Energy Efficiency

Relaying provides advantages when compared to legacy macro/pico base sta-

tions due to wireless backhaul option [44]. Relaying can be also used to boost

the overall network throughput (e.g., urban and suburban scenarios) via effi-

cient utilization of radio resources [20, 21, 27, 33, 55, 70, 70]. The studies done

in [27, 42, 45, 49, 67, 155] show that RN deployment enables macro cell cover-

age extension at the cell edge. It is shown via system level simulations and

field tests [40, 41] that relay UE (RUE) may also experience good throughput

in indoor environments [13, 14, 40, 41, 49].

Furthermore, nomadic relaying is a viable solution for enabling temporary

cellular coverage in special events (e.g., game events, conferences, festivals,

etc.) and in emergency situations (e.g., natural disasters, earthquakes, floods,

telemedicine scenarios, etc.) [14, 140, 141, 151, 156]. To that end, contributions

24

Relaying in LTE-Advanced Systems and Beyond

in [13, 14] show that relaying provides system performance gains in terms of

better coverage and capacity in outdoor-to-indoor scenarios. It is shown that

UEs experience less competition for radio resource on AL which leads to good

throughput gains, even though, RUEs may experience high interference from

eNBs.

It is also shown in [67] that relaying enables cost efficient deployment solu-

tion to cellular network operators, simplify site location planning and can be

easy to deploy in massive numbers on street lamp posts. Simulation results

indicate that operators can save more than 30% of deployment cost by using

RNs in coverage limited LTE-A network [67]. In addition, relaying enables

notable performance gains in terms of SNR and SINR in both interference-

limited and noise-limited cases, provided that RN deployment is done via

proper site planning [18,19, 23, 24, 37, 141].

2.4 Relaying beyond LTE-A

2.4.1 PHY/MAC Layer development

We start by recalling that the PHY layer is defined as different transmission

techniques to send the information bits over the air interface. The MAC layer,

on the other hand, describes different flow and multiplexing mechanisms.

In order to satisfy the future requirements of data hungry wireless applica-

tions and services, several efficient physical layer techniques especially from

the relaying aspect, are being proposed for the future mobile systems. For ex-

ample, the wireless transmission in millimeter wave (mmW) frequency bands

is proposed [82,83, 89–92]. Furthermore, other multicarrier techniques in ad-

dition to OFDM are discussed including the Filter Bank Multicarrier (FBMC)

technique [85–88]. The Massive MIMO scheme [92–100] enables enhanced

spectral efficiency and improved coverage. Inband full duplex relaying is also

one of the relaying flavors. There the wireless nodes are able to execute the

transmission and reception operations at the same time and frequency [126].

Finally, several channel sensing and access schemes in combination with re-

laying are being investigated in [117,118,121,122].

Relaying can be utilized in a cooperative fashion to enable additional com-

munication routes for the data between source and destination. This may

take place through e.g. cooperative diversity or network coding [104, 178].

In cooperative diversity, the data may be exchanged between BS and UE via

25

Relaying in LTE-Advanced Systems and Beyond

two paths, i.e., through the direct link and via the RN [129]. Moreover, the

relay selection mechanisms of opportunistic cooperation also increase the at-

tractiveness of RN employment in the future wireless technologies [179,180].

Opportunistic cooperation enables an option to select a RN which provides

the best communication link between the source and destination [110–120].

Similarly, two-way relaying combined with the wireless network coding can

be exploited in future systems to efficiently utilize the available spectrum

[101–109].

2.4.2 Deployment and Usage Scenarios

The inband relaying as compared to out-of-band relaying, is considered to be

spectral efficient due to self-backhauling capability of RL. Self-backhauling

ability provides advantages for the relaying in upcoming future wireless com-

munication technologies. For example, the RL may employ UDN infrastruc-

ture [6]. In UDN, a large number of small cells are deployed in a given geo-

graphic area (e.g., > 103small cells/km2 [181]). To that end, the wireless back-

hauling (especially inband relaying) is an efficient and rapid option in terms

of deployment cost as well as efficient single frequency utilization [142].

Relaying as a part of UDN deployments can exploit the mmW frequency spec-

trum due to the availability of contiguous large blocks of frequency band in

ranges of 30-300 GHz. The usage of mmW for the small cell operation is feasi-

ble but challenging due to the fact that propagation at high frequency is highly

affected by the physical obstruction. Small cell operation on mmW enhances

the system efficiency as well as increases the capacity in a local coverage area.

The self-backhauling capability also makes relaying a good candidate in small

cell deployment scenarios to enable outdoor-to-indoor coverage in dense ur-

ban area [3].

Mobile and nomadic RNs may make the future communication networks

highly dynamic and make it possible to provide better on-board cellular ser-

vices [151, 152]. To that end, the moving/nomadic relaying, even though

adding system complexity and limitations, are envisioned to be a promising

notion in future. Due to usage of wireless backhaul link, moving RNs in e.g.,

buses and trams, can provide wireless connectivity to UEs in close proximity

as well as enhance the data rate experience of UEs. Moreover, it can also mini-

mizes the penetration losses when the RN enables outdoor-to-indoor cellular

coverage. The nomadic nature of relaying can also provide rapid deployment

flexibilities to network operators on demand basis [153] and is envisioned to

26

Relaying in LTE-Advanced Systems and Beyond

play an important role in future communication technologies [141].

Very recently a drone-based relaying mechanism have been proposed in [154]

to mitigate the traffic congestion in a cellular network as well as enhance the

cellular user performance experience. This work proposes a mechanism to

employ the drone RN in hot spots of 5G cellular networks where the network

is unable to support sudden large traffic volume. Hence, RN are envisioned

to be in nomadic mode in the air, providing the cellular services to users on

behalf of macro cell with improved radio conditions as well as minimum in-

terference. This type of relaying can be also efficiently utilized in scenarios of

emergency events or public safety cases.

The D2D communication may also support relaying to enable the wireless

connectivity between two devices rather than connecting them via infrastruc-

ture BS. This helps the network to separate the local traffic operation from the

global traffic and it allows a direct communication between the devices [143].

This approach may facilitates higher spectral and power efficiency and poten-

tial decreases the system cost and latency [1]. Relaying can be also utilized in

D2D [143] assisted communication with purpose of coverage extension espe-

cially in public safety scenarios [81,144,145]. In D2D relaying, a mobile device

can act as a RN to enable transmission towards mobile devices in the outage

areas [143]. In [146], D2D relaying was used to improve performance in terms

of enhancing the cell edge UE throughput.

Relaying node also bears a potential role in MTC. Here, the BS is controlling a

large number of wireless sensors/devices. Accordingly, congestion problems

on the AL may occur. Hence, RN deployed between the BS and sensors can

decrease the access signaling burden on the BS. To that end, RN aggregates

the traffic received from the sensor devices located in its coverage area and

forward it to the controlling BS [3]. Similarly, the utilization of machine-type

UE relaying extends the coverage area and enable communication between

devices and BS [148,149]. It also increases the battery life of the devices [150].

2.4.3 Relay Backhaul Enhancement Issues

The wireless backhaul plays a major role in the REC network. In literature,

different approaches have been investigated to enhance the backhaul per-

formance. To that end, the RN site planning enables the RN to connect to

the eNB with the best RL and/or smallest shadow fading. The RL limi-

tations are being investigated from the network planning and optimization

perspective in [14, 18–20, 22–24, 28, 33, 37, 70, 140, 141]. Moreover, RL limita-

27

Relaying in LTE-Advanced Systems and Beyond

tions are further relaxed from the perspective of different RRM related issues

in [20, 21, 25–27, 29, 55, 70, 70]. The MIMO/Beamforming/Interference miti-

gation aspects are considered in [12, 13, 30–33], elevation of RL antenna and

optimal path loss models are discussed in [36–39]. Moreover, the RL limita-

tions are further relaxed via CoMP Scheme in [34, 35] and some other initia-

tives based on the flexible backhaul are provided in [182]. The contribution

of [12, 13] analyze the backhaul RL limitations due to interference received

from the neighbouring eNBs. To that end, the performance of transmit beam-

forming and interference mitigation schemes to relax the RL limitation by

minimizing the interference from dominant interfering eNB is investigated.

Results show enhanced system performance on RL in terms of SINR per PRB

and e2e UE throughput. Furthermore, the UL performance of a LTE REC net-

work have been examined in [33, 53, 54, 58, 70, 71].

Table 2.2. Relay Backhaul Link Enhancements in the Literature

References Strategies Study Methods

[14, 18–20,22–24,28, 33,

37, 70, 140,141, 182]

Network

Planning/Optimization

& Nomadic

Analytical and/or

Simulations

[20, 21, 25–27,29, 55, 70,

70]RRM Schemes Simulations

[12,13,30–32,35,35,183] CoMP & Beamforming Simulations

[36, 38, 39] RL Path Loss ModelingSimulations & Field

measurements

2.4.4 Future Research Directions

This section briefly describes some of the open issues and future challenges

occurring in the relay systems. The work in [73] highlights the RRM issues

in multi-hop relaying especially from the call admission control and QoS per-

spective. In [74], authors suggested to address the issue of the enhancements

of self-interference cancellation, reduction of Bit Error Ratio (BER) and Packet

Loss Ratio (PLR) in Full Duplex relaying. In addition to self-interference can-

cellation ability, authors in [76] analyzed the FD relaying in mm-Wave fre-

quency ranges from Energy Efficiency (EE) perspective. The introduction of

novel multi carrier modulation technique known as Index Modulation for

OFDM subcarriers, is one of the future aspects. The aim being to make In-

dex Modulation compatible with the cooperative relay communication sys-

28

Relaying in LTE-Advanced Systems and Beyond

tems [77]. In [75], self-sustainable relaying is envisioned to be one of the ini-

tiatives of the future green wireless communication technologies. The author

in [78] evaluates different buffer-aided RN selection algorithms and draws a

broad picture of future challenges to be addressed from the perspective of

buffer-aided relaying in 5G communications. A survey done in [79] discusses

on the role of different relaying strategies in future LTE including Type-1,

Type-2, and moving RN. Figure 2.2 presents the references regarding the fu-

ture challenges of the relaying system.

29

Relaying in LTE-Advanced Systems and Beyond

30

3. Resource Optimal Relaying

3.1 Background

In this chapter, we analyse the performance of dual-hop DF relaying in terms

of e2e data rates when the radio resources are shared between RL and AL

using different Resource Allocation (RA) schemes. While the focus is on the

conventional case where DL and UL are carried out with the same DeNB,

we also briefly consider the impact of different RA schemes in the context of

relaying systems where UEs can decouple DL and UL connections. In what

follows, we apply a simple resource allocation model where only a single user

is considered. We note that this model simplifies the analytical study but prin-

ciples of the proposed resource allocations can be extended to the multi-user

scenarios. Such extension would, however, lead to unnecessary complex anal-

ysis and is not considered in this thesis.

3.2 Previous work and Contributions

3.2.1 Previous work

Traditionally, the performance analysis of dual-hop DF relaying has been con-

ducted in terms of ergodic capacity [184–188], outage capacity [189,190], out-

age probability [191–193], and the bit error probability [193]. In [184], author

studied the ergodic capacity of DF relaying in the presence of Rician fading

and, based on the obtained results, concluded that the combination of the

optimal power control and rate adaptation provides a notable positive effect

on the overall system performance. In [185] and [186], closed form expres-

sions were derived for the ergodic capacity of DF relaying assuming different

31

Resource Optimal Relaying

approaches. For example, Taylor’s series expansions were used in [185] to

determinel the lower and upper bounds for the ergodic capacity, while the

study carried out in [186] focused on the ergodic capacity when indepen-

dent non-identically distributed composite inverse Gaussian/Nakagami-m

random variables were used to model the wireless channel. The research work

done in [189] evaluated the benefits of power allocation on the outage capac-

ity of dual-hop DF relaying. Authors showed that the gain of optimal power

allocation is more pronounced when channel power gain difference grows

between RL and AL. In [190], the outage capacity formula for a dual-hop DF

relaying system was derived when different generalized fast fading channel

models were used.

In [187, 191–193], the performance gain that dual-hop DF relaying provides

when communication takes place in parallel with the direct communication

between eNB and UE is characterized. Authors of [187] studied the ergodic

capacity of DF relaying in presence of independent but non-necessarily iden-

tically distributed Nakagami-m fading. Similarly, the average symbol error

probability and outage probability was calculated in [191] by studying the in-

stantaneous received e2e SNR of dual-hop DF relaying in case of Nakagami-m

fading. Moreover, the authors of [192, 193] obtained closed form expressions

for the outage and bit error probability of DF relaying in presence of dis-

similar Rayleigh fading channel models. The ergodic capacity analysis was

extended in [188] for the multi-hop DF relaying scenario, where two upper

bounds were obtained using the Jensen’s inequality and by computing the

harmonic-geometric means in case of a Rayleigh fading. Finally, the authors

of [131] performed a comparative study between the dual-hop regenerative

(i.e., DF) and the non-regenerative (i.e., AF) relaying systems, by deriving

closed form expressions for the statistics of SNR harmonic means of both

links. Moreover, the impact of buffering on the performance of DF relaying

has been investigated in [194]. Also, different authors have studied the re-

laying performance in terms of enhanced spectral efficiency by proposing the

implementation of rate adaptation and relay selection scheme [195], as well

as adaptive modulation schemes [196] while the performance and energy ef-

ficiency of REC networks has been investigated in e.g. [19, 42, 197–205].

It is worth noticing that in all these previous works, the communication re-

sources of the dual-hop relaying system are equally shared between RL and

AL. Though, some recent work [195, 196] considered the optimal RA scheme

introduced in [45]. This study recall the said optimal RA scheme where the

same amount of data is transferred over both RL and AL at each radio frame.

32

Resource Optimal Relaying

According to our best understanding, analysis of e2e data rate and the deriva-

tion of closed-form expressions for its distribution and expected value in pres-

ence of optimal resource allocation have not been well addressed before our

publication [10]. Part of the material of this chapter will be also published

in [11].

3.2.2 Contributions

The main contributions of this chapter are the following: First, we deduce

closed-form expressions that describe the performance impact of the resource

allocation on the relaying when the UE is connected to the same RN and DeNB

in DL and UL. Second, we consider the case where the DL/UL communica-

tion between the source and destination via RN, is decoupled, meaning that

the source (destination) node of a DL (UL) transmission can be any macro

eNB that has the target UE in range (via an intermediate RN).

To that end, the first case includes the comparative analysis of several RA

schemes: that is, conventional RA (where the RL and AL equally share the

time resources), fixed RA (where the RL and AL resources are proportionally

allocated according to their corresponding mean data rates), and optimal RA

(where the RL and AL resources are allocated according to the instantaneous

data rates in each link). Moreover, we also derive a closed-form expressions

for the probability distribution function for the mean and outage rate of the

dual-hop DF relaying system, by assuming the optimal RA scheme. While the

mean e2e data rate attains an expression in terms of an integral which does

not admit closed-form solution, so we can deduce a tight lower bound expres-

sion which approximates accurately the mean e2e data rate. Results are being

published in [10].

In case of DL/UL decoupling, we also analyze the e2e data rate performance

when different RNs, while possibly connected to different macro eNBs, are

being employed to retransmit the information in both DL (eNB-RN-UE) and

UL (UE-RN-eNB) direction of communication. Study assumes the adaptive

modulation and coding in both RL and AL, in order to obtain improved spec-

tral efficiency for the given SNR value. The work includes the derivation of

closed form expressions of e2e data rate for both DL and UL. Results will be

published in [11].

Chapter is organized as follows: Section 3.3.1 present the system model of

both coupling and decoupling cases and gives the channel modeling assump-

tions. Section 3.3.2 explains the different RA schemes focusing on the differ-

ences between mentioned conventional RA, fixed RA with/without buffering,

33

Resource Optimal Relaying

and optimal RA. Moreover, Section 3.4.1 provides the performance analysis of

the studied RA schemes. Finally, conclusions are drawn in Section 3.5.

3.3 System Model and Resource Allocation Schemes

3.3.1 System Models

DL/UL coupled scenario

The considered dual-hop relay system is shown in Figure 3.1. We assume that

RN is acting as an intermediate node between UE and eNB and the communi-

cation between both will always be taken via RN. We denote the radio frame

duration by T and assume the radio resources can be shared in time between

RL and AL assuming an infinitesimal granularity. We employ the HD relay-

ing, where RN cannot transmit and receive information simultaneously. We

denote T = Tr + Ta, where Tr and Ta refer to transmission times allocated for

RL and AL, respectively. Moreover, the Rr and Ra denote the instantaneous

data rates on RL and AL in bits per second. We note that block fading model

is assumed where the channel remains constant over each TTI but change be-

tween every TTIs depending upon the channel statistics. Furthermore, the

RL and AL data rates are dependent on the SNR levels and the bandwidths

allocation to both RL and AL. Hence, the achievable data rates during a TTI

on RL and AL are given by Shannon’s formula as

Rr = Wr log2(1 + γr), and Ra = Wa log2(1 + γa), (3.1)

where Wr and Wa refer to the bandwidths allocated to RL and AL, respec-

tively, while γr and γa represent the instantaneous SNR values experienced on

RL and AL, respectively. In fixed relaying, RN is part of the network and RN

locations are defined in the network planning process, so that a very strong

Line-of-Sight (LoS) maybe experienced on RL towards the DeNB. We note

that in 3GPP system scenarios channel models for the RL experience LoS as

well [9]. Due to LoS, there is high likelihood that RL SNR admit only small

variations and hence, Ricean distribution can be applied to model the fading

in RL [206, 207]. On the other hand, a Non-Line-of-Sight (NLoS) condition is

assumed in AL because, a UE typically changing its location due to mobility.

Thus, we can apply the Rayleigh distribution in order to model the fast fading

component of the channel in the AL, as shown in Figure 3.1.

34

Resource Optimal Relaying

RL(Ricean fading)

DonoreNB

RN UE

AL(Rayleigh fading)

Figure 3.1. System model for the fixed relaying system consisting of a eNB, RN, and UE. Fast

fading component of the wireless channel in the RL and AL are modeled according

to a Ricean (LoS) and Rayleigh (NLoS) distribution.

DL/UL Decoupling scenario

In this scenario, we consider a two-cell system model of Figure 3.2, where

RNs are associated to different macro eNBs and serve a common UE located

at the cell edge of both cells. It is possible to enhance the performance in both

communication directions by applying the UL and DL decoupling whenever

it is suitable from the e2e data rate perspective. We assume Frequency Di-

vision Duplexing (FDD) to enable the simultaneous transmission in both UL

and DL directions. Moreover, RNs are HD and the temporal resources are or-

thogonally shared between AL and RL in both directions. No power control

is applied in DL but in UL direction UE uses an open-loop power control al-

gorithm with target received power p(ul)0 [146]. The channels are modelled as

in the aforementioned coupled case.

Donor eNB 1

RN 1

UE

DownlinkServing Cell

Donor eNB 2

RN 2

UplinkServing Cell

Uplink Power Control (P0)

Figure 3.2. System model for a two-cell DF relaying system with UL and DL decoupling. No

power control is implemented in DL. Open-loop power control with target received

power p(ul)0 is used in UL to control the interference.

35

Resource Optimal Relaying

3.3.2 Resource Allocation Schemes

We consider four different RA schemes: Conventional RA, Fixed RA

with/without buffer, and Optimal RA. The conventional RA and fixed RA

represent benchmark for the optimal RA scheme. The reference RA schemes

allocate the communication resources to AL and RL with fixed Ta and Tr.

Conventional RA. In a conventional RA scheme [208, 209], the time resources

are equally divided between AL and RL, i.e., Ta = Tr = 0.5 · T . Thus, the

conventional e2e data rate per unit bandwidth (Rconve2e ) for this RA scheme at a

given fixed transmission power, will be as

Rconve2e =

1

2min{Rr , Ra }. (3.2)

Moreover, the e2e performance depends on the RL and AL SNRs while the

link with lowest SNR will become a bottleneck for the overall e2e data rate.

Here, the factor 1/2 represents the half-duplex constraint.

Fixed RA without buffer. If there is no buffer, then the values of Ta and Tr are

chosen proportional to the mean data rate of each link, i.e., E{Rr} and E{Ra}.

Hence, we write

T Rwobe2e = Ta E{Ra} = Tr E{Rr}. (3.3)

From (3.3), it can be seen that

Ta =Tr E{Rr}E{Ra} , Tr =

Ta E{Ra}E{Rr} . (3.4)

In general E{Rr} �= E{Ra}, and thus the time allocations for each link will

be different i.e., Tr �= Ta verifying Ta + Tr = T . The overall payload data

transmitted over the e2e link is given by

P = min {TaRa, TrRr} . (3.5)

Using (3.5), the e2e data rate can be written as follows:

Rwobe2e =

P

T=

min {TaRa, TrRr}T

,

=1

Tmin

{Tr E{Rr}E{Ra} Ra, TrRr

},

=1

Tr +Tr E{Rr}E{Ra}

min

{Tr E{Rr}E{Ra} Ra, TrRr

}

(3.6)

36

Resource Optimal Relaying

Here, we can apply the formula min{X,Y } = 1/2(X + Y − |Y −X|) in (3.6)

and obtain

Rwobe2e =

E{Ra}Tr (E{Ra}+ E{Rr})

1

2

[Tr E{Rr}E{Ra} Ra + TrRr −

∣∣∣∣Tr E{Rr}E{Ra} Ra − TrRr

∣∣∣∣]

=1

2

E{Ra}E{Ra}+ E{Rr}

[E{Rr}E{Ra}Ra +Rr −

∣∣∣∣E{Rr}E{Ra}Ra −Rr

∣∣∣∣].

(3.7)

From (3.7), three different e2e data rate formulas are obtained by solving the

modulus operator, i.e.,

Rwobe2e =

⎧⎪⎪⎪⎪⎨⎪⎪⎪⎪⎩

Rr E{Ra}E{Rr}+E{Ra} ,

E{Rr}E{Ra}Ra > Rr,

Ra E{Rr}E{Rr}+E{Ra} ,

E{Rr}E{Ra}Ra < Rr,

12Rr E{Ra}+Ra E{Rr}

E{Rr}+E{Ra} , E{Rr}E{Ra}Ra = Rr.

(3.8)

Fixed RA with buffer. In fixed RA scheme with buffer, RNs have large buffers

which avoid data overflow during the transmissions. In buffer-aided relay-

ing, RN can better use the CSI while selecting between reception in RL and

transmission in AL. RN can also utilize the buffer effectively in scheduling.

In buffer-aided relaying, the eNB can continue the transmission on the RL

without taking into account of the AL channel quality. When the AL is in

deep fade, the data can be stored at the RN’s buffer until the AL channel con-

dition becomes suitable for transmission again. Then, the RN can forward

the previously buffered data to UE, provided that the AL channel conditions

are favorable [3, 78, 84]. We denote the e2e data rate by Rwbe2e for this relaying

setup. Though the same amount of data is transferred over both RL and AL

in long-term, the relation

T · E{Rwbe2e} = Ta · E{Ra} = Tr · E{Rr} (3.9)

should hold. Then, we find that

E{Rwbe2e} =

Tr · E{Rr}Tr + Ta

=E{Rr}

1 + Ta/Tr

=E{Rr}E{Ra}

E{Rr}+ E{Ra} .(3.10)

Optimal RA. The optimal resource sharing notion for the L-hops connections

has been considered previously in [45]. In optimal RA scheme, the e2e data

rate Ropte2e is maximized, which happens when the same amount of data is

transferred over each link, i.e., RL and AL, at each TTI. This is equivalent to

T Ropte2e = TrRr = TaRa. (3.11)

37

Resource Optimal Relaying

To obtain the condition (3.11), it is required to properly allocate the radio re-

sources among RL and AL. We note that the selection of Tr, Ta, Wr, and Wa

depends on the instantaneous SNR of each link. To that end, one degree of

freedom is to adjust the transmission power of RN. However, this study con-

siders the LTE-A style fixed relaying, where the transmission power of eNBs

and RNs in DL, is kept constant in order to prevent the potential harmful co-

channel interferences toward the adjacent cells. Hence, we are left with time

and bandwidth resources to obtain (3.11). Implementation of resource opti-

mal relaying (such as Type 2 DF relaying in LTE-A and Mobile WiMAX [210])

require flexibility to allocate the radio resources in time-frequency domain,

so that optimal resourcing can be achieved, or at least, well approximated.

Using (3.1) and (3.11), the expression for the optimal e2e data rate can be ob-

tained as follows:

Ropte2e =

Tr Rr

Tr + Ta=

Rr

1 + Ta/Tr=

Rr

1 +Rr/Ra=

Rr Ra

Rr +Ra. (3.12)

Performance comparison of the different RA schemes. While, comparing the effi-

ciencies of the mentioned RA schemes, we find that

Rmax ≥ E{Rwbe2e} ≥ E{Ropt

e2e} ≥ E{Rwobe2e } ≥ E{Rconv

e2e } (3.13)

It is known that in conventional RA scenario, RN does not have a buffering

capability and thus, it always equally divides the communication resource

between RL and AL. Obviously, performance of conventional RA is clearly

inferior to the rest of the RA strategies. Still, the conventional RA is the sim-

plest scheme to distribute the communication resources in practice. Fixed RA

without buffer allocates the time resources of the radio frames to fit with the

long-term (expected) data rates of each individual link i.e., RL and AL, as

given by (3.3). Though, the e2e data rate of this relaying RA scheme (having

no buffering capability) is inferior to the fixed RA scheme with buffer as well

as to the optimal RA scheme that adjusts the transmission times and data

rates in RL and AL without delay, see (3.11). Furthermore, it is noted that

the relaying system with buffer ability, outperforms the resource optimal RA

scheme as given in (3.12) when expected e2e data rates are compared. While it

can be Ropte2e > Rwb

e2e for instantaneous data rates, in fixed RA with buffer it is

possible to set Rr = E{Rr} and Ra = E{Ra}. Note also that Jensen’s inequal-

ity verifies E{Rwbe2e} ≥ E{Ropt

e2e}, see (3.12) and (3.13). However, it is worth to

mention that the buffer usage leads to additional latency as well as additional

hardware costs, as the data blocks transported over the RL and AL of a radio

frame do not usually match over the same TTI.

38

Resource Optimal Relaying

Figure 3.3. Cumulative Distribution Function of the instantaneous gains that resource optimal

RA provides with respect to the other RA schemes in a dual-hop DF relaying. Pa-

rameters: Wr = Wa = 180 kHz, E{γr} = γr = 15dB (Ricean K-factor = 12 dB),

and E{γa} = γa = 5dB.

Figure 3.3 shows the CDF of the gains that can be achieved by means of

the resource optimal RA with respect to the above mentioned baseline RA

schemes (i.e., conventional RA and fixed RA with and without buffer). It is

observed that the Optimal RA always enables significant performance gains

when compared to conventional and fixed RA without using buffer. We recall

that in case of the fixed RA with buffer, it is possible to set Rr = E{Rr} and

Ra = E{Ra}; hence, sometimes Ropte2e < Rwb

e2e, as shown in Figure 3.3. Yet, Ropte2e

represents the rate-optimal RA scheme provided that there is no buffer at RN.

3.3.3 Role of the Channel State Information

We assume that the DL receiver has always short-term CSI which is obtained

from pilot signals through a suitable channel estimation process, while, trans-

mitter may have either long-term or short-term CSI only. Hence, the receiver

enables the CSI via a certain feedback mechanism. Feedback is fast for short-

term CSI and slow for long-term CSI. Here, ’fast’ and ’slow’ denote the speed

of the feedback provision when compared to the channel coherence time.

It is assumed that a RA scheme applying instantaneous rates (i.e. Rr and Ra)

should have short-term CSI of both links, i.e., RL and AL, in order to take the

RA decisions per TTI. This short-term CSI is needed in the resource optimal

RA scenario. Here, the RN should request CSI report from the UE before per-

forming the allocation of resources (jointly with DeNB) on RL and AL. More-

over, in case of fixed RA without buffer, the communication time resources,

Tr and Ta are chosen proportional to the mean data rate of RL and AL. Thus,

there is only need of long-term CSI to allocate the communication resources

which, after being determined, are kept fixed for all TTI. In the same way,

fixed RA with buffer also require only long-term CSI. Finally, we recall that

39

Resource Optimal Relaying

if fast radio layer scheduling is applied in a radio link, then short-term CSI is

always needed.

3.4 Performance analysis

3.4.1 DL/UL coupled scenario

In the following, the CDF of the e2e data rate is obtained and, then, closed

form expressions are deduced for both the mean and the outage rates.

SNR distribution in the Relay Link

In case of resource optimal RA, the e2e data rate is obtained by combin-

ing (3.1) and (3.12) as follows:

Ropte2e =

Wr log2(1 + γr)Wa log2(1 + γa)

Wr, log2(1 + γr) +Wa log2(1 + γa), (3.14)

where γr and γa are the instantaneous SNRs of the RL and AL, respectively.

If the fast fading components of these links admit Ricean (LoS) and Rayleigh

(NLoS) fading distributions, then the Probability Density Functions (PDFs)

for γr and γa are given by [207]:

fγr(γ) =K + 1

γre−Kγr+(K+1)γ

γr I0

(2

√K(K + 1)γ

γr

)(3.15)

and

fγa(γ) =1

γae− γ

γa . (3.16)

Here γr and γa refer to the mean SNR in RL and AL. Moreover, in (3.15),

I0 is the zero order modified Bessel function of the first kind [211], which

makes the mathematical analysis of (3.14) challenging. Nevertheless, in case

of strong LoS scenarios, the Ricean K-factor can be large, as noted in [206],

where typical value was of the order 12dB. Presuming a high K-factor, the RL

SNR is almost constant in LoS condition. Although results of [206] enables a

good justification for this assumption, we have also verified it via simulations:

40

Resource Optimal Relaying

Figure 3.4. Accuracy of the constant SNR assumption on RL. The CDF of the Re2e is presented

here for different combinations of mean SNR in both RL (γr) and AL (γa).

The CDF of Ropte2e in (3.14) is presented by Figure 3.4, where the AL SNR is

considered to always distributed according to (3.16). Furthermore, we com-

pare two scenarios: First (solid curves), it is assumed that fast fading in the

RL is distributed according to (3.15). Second (dashed curves), we assume a

constant RL SNR, i.e., (γr = γr).

Figure 3.4 presents the CDF of Ropte2e for different values of γr and γa, with

K = 12dB in all cases. It is observed that the assumption of constant SNR is

valid not only in terms of the resulting distributions, but also in terms of the

deviation of the average Ropte2e , where for values K > 12 dB it was verified that

the approximation error is always smaller than 1%. We also note that using

the Jensen’s inequality we obtain

E{Rr} = E{log2(1 + γr)} ≤ log2(1 + E{γr}), (3.17)

where the lower bound for the mean e2e data rate becomes tight as the K-

factor of the Ricean distribution grows.

Probability distribution for the e2e data rate

This subsection presents the e2e data rate distributions, that is CDF FRopte2e

and PDF fRopte2e

, when resource optimal RA scheme is employed. We note

that the SNR on the RL is assumed to be constant while the AL SNR - de-

noted by γa - admit uncorrelated values from TTI to TTI. From (3.12), the

CDF FRopte2e

can be deduced using the standard mapping procedure between

random variables [212]. First, we set re2e = g1(Ra) = Rr · Ra/(Rr + Ra) and

after solving the instantaneous AL data rate Ra in terms of re2e, we obtain

g−11 (re2e) = re2e ·Rr/(Rr− re2e). The CDF of Ropt

e2e admits the following formu-

lation:

FRopte2e

(re2e) = FRa(g−11 (re2e)) = FRa

(re2e ·Rr

Rr − re2e

). (3.18)

41

Resource Optimal Relaying

Second, the SNR in the AL can be expressed as a function of γa as follows;

γa = g−12 (ra) = e(

ln 2Wa

·ra) − 1. Then, it is possible to write the CDF of Ra as:

FRa(ra) = Fγa(g−12 (ra)) = Fγa

(e

ln 2Wa

·ra − 1). (3.19)

After combining (3.12), (3.18) and (3.19), FRopte2e

is given by

FRopte2e

(re2e) = Fγa

(e

ln 2Wa

re2eRr(Rr−re2e) − 1

), (3.20)

where FRopte2e

(re2e) = 0 if re2e < 0 and FRopte2e

(re2e) = 1 if re2e ≥ Rr. Thus,

using (3.16), FRopte2e

becomes

FRopte2e

(re2e) = 1− e− 1

γa

(eln 2Wa

re2eRr(Rr−re2e)−1

). (3.21)

Using (3.20) we can deduce CDF for e2e data rate Re2e also in case AL SNR

admit e.g. Nagakami or Weibull distributions. The probability density fR(r)

of e2e data rate can be obtained by taking the derivative of CDF (3.21):

fRopte2e

(re2e) =ln 2

γaWaexp

⎛⎝−

exp(

ln 2Rr re2eWa (Rr−re2e)

)− 1

γa

⎞⎠

× R2r

(Rr − re2e)2 exp

(ln 2Rr re2e

Wa (Rr − re2e)

).

(3.22)

The e2e data rate PDF fRopte2e

and CDFFRopte2e

are illustrated by Figure 3.5 and 3.6,

respectively. Here, in Figure 3.6, the mean SNR on the RL (γr) is assumed to

be 20 dB. The dashed vertical line on the right-hand side of the figure denotes

the upper limit for the e2e data rate. This upper limit is obtained if the SNR

on AL grows very large (i.e., upper bound for Ropte2e). It also includes the CDF

curves for AL SNR of 0 dB (black curves), 10 dB (orange curves) and 20 dB

(green curves).

Results show the impact of bandwidth allocation between RL and AL. By as-

suming Wa/Wr = 1 (dotted curves), and Wa/Wr = 2 (dashed curves) and

Wa/Wr = 4 (solid curves), we find that the distribution of re2e tends asymp-

totically to Rr with increasing values of γa and Wa/Wr. It can be seen from

results that the imbalance between the mean SNR on the RL and AL signif-

icantly deteriorate the e2e data rate performance even though the resource

optimal RA scheme is applied. If the mean SNR in AL is considerably lower

than the mean SNR in RL, then this imbalance can be compensated by allo-

cating more frequency resources to the AL (see black curves).

42

Resource Optimal Relaying

0 2 4 6 8 10

105

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1 10-5

Figure 3.5. PDF of e2e data rate when SNR γr on the RL is 20 dB and Wr=180 kHz.

0 2 4 6 8 10

105

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Figure 3.6. CDF of e2e data rate when SNR γr on the RL is 20 dB and Wr=180 kHz.

It is noticed that for Wa/Wr = 4 the system gains for γa=10 dB and γa=20 dB

as compared to γa=0 dB at 10%-ile, 50%-ile, 90%-ile are 383%, 107%, 45%, and

750%, 153% and 62%, respectively. Moreover, in case of low mean SNR in

AL, the allocation of additional frequency resources to AL doesn’t lead to any

notable enhancement in the e2e data rate (see solid black curve).

From Figure 3.6 we find that if the mean SNRs in RL and AL are the same (see

green curves), then the e2e data rate is significantly enhanced in both low

and high probability regions when increasing the frequency resources in AL.

For γa = 20 dB, the system performance gain for Wa/Wr = 2 and Wa/Wr = 4

compared to Wa/Wr = 1 at 10%-ile, 50%-ile, 90%-ile are 49%, 35%, 29%, and

43

Resource Optimal Relaying

96%, 63% and 56%, respectively. Hence, it can be said that increasing the

resources in AL, will results a positive impact on e2e performance when the

RL and AL SNR imbalance is considerably small.

It is observed that even though the AL resources are four times larger than in

RL the e2e data rate is not very close to the performance upper bound. The

e2e data rate of the resource optimal RA scheme asymptotically approaches

the upper limit when Wa/Wr increases. Yet the increased e2e data rate effi-

ciency from additional frequency resources in AL decreases when the Wa/Wr

becomes large.

Mean e2e data rate for the resource optimal RA scheme

This subsection presents the mean (average) data rate Cav for the resource

optimal data rate adaptation over the e2e link as given by computing the ex-

pected value of Ropte2e :

Cav � E{Ropte2e} =

∫ ∞

0r fRe2e(r) dr. (3.23)

We note that this definition of mean data rate is well in line with the con-

ventional definition of mean data rate for a single-hop link as given in [213].

Equation (3.23) can be written in the form

Cav =

∫ ∞

0

Rr · raRr + ra

fRa(ra)dra =

∫ ∞

0

Rr ·Wa log2(1 + γa)

Rr +Wa log2(1 + γa)fγa(γa) dγa. (3.24)

Thus mean data rateCav is computed for fixed AL bandwidthWa and RL data

rate Rr.

At this stage we note that since τr = Tr/T = Ropte2e/Rr, the expected time shar-

ing for RL is given by E{τr} = Cav/Rr. This value can be used if time sharing

between RL and AL needs to be set. In the following we consider the com-

putation of Cav in the case of exponentially distributed AL SNR. Since CDF

of Ropte2e is easier to handle than the PDF, we use integration by parts to obtain

the form:

Cav =

∫ ∞

0

(1− FRopt

e2e(r)

)dr.

Then, we apply (3.21) and the following substitution:

r =Rr · Wa

ln 2 ln t

Rr +Waln 2 ln t

, dr =R2

r · Waln 2 · dt

t(Rr +

Waln 2 ln t

)2 .Since limr→0 t(r) = 1 and t(r) → ∞ when r → Rr, Cav can be re-written as

follows:

Cav = R2r ·

Wa

ln 2

∫ ∞

1

e−(t−1)/γadt

t(Rr +

Waln 2 ln t

)2 . (3.25)

After integration by parts, we can write (3.25) in the form

Cav = Rr

(1− 1

γae1/γa

∫ ∞

1

e−t/γadt

1 + WaRr ln 2 ln t

). (3.26)

44

Resource Optimal Relaying

Since closed form expression for the integral in (3.26) does not exist, we will

later deduce a tight lower bound for the Cav which can serve also as a good

approximation. Let us consider next the distribution for the relative time

(τr = Tr/T ) to determine the time resource allocation for the RL transmission.

According to (3.11), we have τr = R/Rr and

Fτr(τ) =1

RrFR

Rr

). (3.27)

The time sharing τa = Ta/T for AL can be computed from the equality τa =

1−τr. The CDF presented in (3.27) can be used in the network planning phase

for the dimensioning. This requires that the relative transmission time used

in the RL is under some predefined threshold τmaxr with a given probability

Pmaxr . This requirement will help to save some communication resources in

RL, in order to enable service to those UEs that are directly connected to DeNB

on the direct link.

It is noted that (3.27) also indicates the relative data rate loss when compared

to the ideal RL. The loss in the e2e data rate when compared to the ideal RL,

is Ra − R, and the relative loss is δR = (Ra − R)/Ra = R/Rr. This relative

data rate loss can be utilized as a parameter to compare the data rate perfor-

mance between RN (with data rate limitation in the RL) and pico base station

(without data rate limitation in the backhaul).

Outage rate for the resource optimal RA scheme

The outage rate is given as a largest data rate Rout such that P (R < Rout) =

P out is verified, where P out is defined as the outage probability [213]. In order

to derive a mathematical expression for the data rate Rout, we write

P out = P (Ropte2e < Rout) = P

(Ra ·Rr

Ra +Rr< Rout

)

= P

(Ra <

Rr ·Rout

Rr −Rout

)= Fγa

(e

ln 2Wa

· Rr·Rout

Rr−Rout − 1

),

(3.28)

where the latter equality follows from (3.20). The outage rate Rout can be

computed from (3.28) analytically, provided that the inverse of Fγa(γ) admit

closed-form expression. If AL SNR follows the exponential distribution, then

we find that

Rout =Rout

a ·Rr

Routa +Rr

=Waln 2 ln

(1− γa ln

(1− P out

)) ·Rr

Waln 2 ln (1− γa ln (1− P out)) +Rr

,

(3.29)

where Routa is the outage rate obtained from

Routa = Wa · log2(1 + γout

a ), Fγa(γouta ) = P out. (3.30)

45

Resource Optimal Relaying

Hence, from (3.29) and (3.30), we see that e2e data rate formula (3.12) defines

a direct relation between the outage rate in AL and e2e link. Here we recall

the discussion on the time sharing between RL and AL. It is noted that using

the distribution (3.27), we can similarly with (3.29) compute the time sharing

threshold τmaxr for a given excess probability Pmax

r .

Figures 3.7 and 3.8 present the mean e2e data rate Cav and outage rate Rout,

respectively as a function of mean AL SNR (γa). Here, the mean SNR in RL

is 20 dB. The black horizontal dotted line denotes the upper limit for Cav and

Rout. It is shown that both parameters increase with the increase of mean AL

SNR γa.

-5 0 5 10 15 20 25 30 35 40 45 50γa [dB]

0

2

4

6

8

10

12

Average

e2erate

Cav

[bps]

×105

Wa= Wr

Wa= 2Wr

Wa= 4Wr

Rr

Figure 3.7. Mean e2e data rate as a function of AL SNR γa when SNR γr on the RL is 20 dB and

Wr=180 kHz.

It is observed from Figure 3.7 that the mean e2e data rate with resource op-

timal RA scheme is significantly enhanced when employing additional fre-

quency resources in the AL. We consider two example cases: Cav = 0.25Mbps

and Cav = 0.75Mbps. We note that Cav = 0.25Mbps is obtained when

γa = −1dB for Wa/Wr = 4, and γa = 7dB for Wa/Wr = 1. However, if

we target to the mean data rate of 0.75Mbps, then the mean SNR in the AL

should increase to γa = 12dB for Wa/Wr = 4, and γa = 40dB for Wa/Wr = 1.

46

Resource Optimal Relaying

-5 0 5 10 15 20 25 30 35 40 45 50γa [dB]

0

2

4

6

8

10

12

Outage

e2erate

Rout[bps]

×105

P out = 5%, Wa= Wr

P out = 0.5%, Wa= Wr

P out = 5%, Wa= 2Wr

P out = 0.5%, Wa= 2Wr

P out = 5%, Wa= 4Wr

P out = 0.5%, Wa= 4Wr

Rr

Figure 3.8. Outage e2e data rate as a function of AL SNR γa when SNR γr on the RL is 20dB

and Wr=180 kHz.

Similarly, Figure 3.8 presents the outage rate for the outage probability 5%

(solid curves) and 0.5% (dashed curves), respectively. As it is expected, the

outage data rate Rout decreases when decreasing the outage probability. It

can be also seen that as the mean SNR in the AL increases, the dashed and

solid curves with the same color come closer to each other. By this, it can be

concluded that the performance loss for guaranteeing a better e2e radio link

reliability (lower outage probability) can be partly compensated by allocating

more resources to AL. It is also observed that the impact of a larger bandwidth

allocation is more prominent when the AL SNR is low.

Lower bound for the average e2e data rate

The closed-form expression for (3.26) does not exist. In order to obtain the

computable expression, the integration domain in (3.26) is divided into three

sub-intervals such that

I =

∫ ∞

1

e−t/γadt

1 + WaRr ln 2 ln t

= I1 + I2 + I3,

where the corresponding integration domains are (1,mγa), (mγa,Mγa) and

(Mγa,∞), m < 1 < M . The rationale behind this approach is that each inte-

gral Ix with x = {1, 2, 3} admit a closed form bound. Exact expressions for

bounds are derived below.

For the first integral we have

I1 <N∑

n=0

(−1)n

n!

(1

γa

)n ∫ mγa

1

tn

1 + WaRr ln 2 ln t

dt, (3.31)

where N is odd. Thus, we replace in integral of (3.26) the exponential term by

47

Resource Optimal Relaying

its Taylor expansion. We note that (3.31) provides an upper bound only when

the number of terms in the sum is odd. The right side function in (3.31) can

be integrated to obtain a closed-form expression as given by (3.33). We start

computation by writing∫ mγa

1

tn

1 + b ln tdt =

∫ mγa

0

tn

1 + b ln tdt−

∫ 1

0

tn

1 + b ln tdt. (3.32)

In order to find the expression in terms of exponential integral function, we

substitute s = tmγa

and dt = mγads. Note that s = 0mγa

= 0 for t = 0 and

s = mγamγa

= 1 for t = mγa. Hence, integrals can be rewritten as

∫ 1

0

(mγas)n

1 + b ln(mγas)mγa ds−

∫ 1

0

tn

1 + b ln tdt

=(mγa)

n+1

b

∫ 1

0

sn

(1b + ln(mγa)) + ln sds−

∫ 1

0

tn

1 + b ln tdt

To further simplify the above expressions, we apply∫ 10

xp−1

q + lnxdx =

e−pqEi(pq) given in [211, Prop. 4.281.3]. Then, after tedious but straightfor-

ward computation we obtain

I1 <N∑

n=0

(−1)n

n!

(1

γa

)n e−nb

b

(Ei(n ln(e

1bmγa))− Ei(

n

b)). (3.33)

The upper bound for the integral I2 can be calculated by assuming the lin-

ear interpolation Ψ(t) on interval (mγa,Mγa) such that Ψ(mγa) = ln(mγa),

Ψ(Mγa) = ln(Mγa). Then Ψ(t) < ln t on (mγa,Mγa) and we have

I2 <

∫ Mγa

mγa

e−t/γadt

1 + bΨ(t)=

∫ Mγa

mγa

e−t/γadt

1 + b(pt+ q), (3.34)

where

p =ln(M/m)

(M −m)γa, q = ln(mγa)−

(m

M −m

)ln(M/m).

Exploiting p, q and the exponential integral function from [214] as given∫ eCZ

AZ +B= 1

AEi

(C(BA + Z)

)e−

BCA , integral in (3.34) can be rewritten in

the form

I2 =

(k2 γab k1

)· e

⎛⎝k2 + bM ln(mγa)− bm ln(Mγa)

b k1

⎞⎠

×(E1

(k2

1 + b ln(mγa)

b k1

)− E1

(k2

1 + b ln(Mγa)

b k1

)).

where k1 = ln (M/m) and k2 = M −m.

Finally, we use a simple bound in the interval (M γa,∞) as given:

I3 <

∫ ∞

M γa

e−t/γa dt

1 + WaRr ln 2 ln(Mγa)

=γa e

−M

1 + WaRr ln 2 ln(M γa)

. (3.35)

48

Resource Optimal Relaying

This formula is accurate when Mγa >> 1. Combining the above results we

obtain a lower bound for Cav:

Cav > CLB = Rr ·(1− 1

γae1/γa(I1 + I2 + I3)

). (3.36)

0 2 4 6 8 10 12 14 16 18 200.2

0.4

0.6

0.8

1

0 2 4 6 8 10 12 14 16 18 200

5

10

15

Figure 3.9. Upper: Mean e2e data rate as a function of the SNR in the AL and the lower-bound

given by (3.36). Lower: Accuracy of (3.36), error = 100 Cav−CLBCav

. Optimization for

γa ∈ [0, 20]dB: m = 0.995 and M ∈ [2.42, 2.61].

Figure 3.9 shows that the accuracy of CLB increases rapidly with γa, with an

error smaller than 2% for γa > 2dB, which is the range of interest from a

practical point of view. The applied m and M has been derived via numerical

methods.

3.4.2 DL/UL decoupling scenario

Here we analyse the e2e data rate of a dual-hop relaying for the case when DL

and UL are decoupled. This discussion is based on the work in [11]. Like in

the coupled case, we assume a block fading model and recall the (3.5) which

gives the e2e data rate for the dual-hop relaying system:

Re2e T = min{Ra Ta, Rr Tr}. (3.37)

We consider two RA schemes to distribute the communication resources be-

tween RL and AL. First, Fixed long-term RA scheme scheme allocates resources

in each direction on long-term basis such that mean e2e data rates in both RL

49

Resource Optimal Relaying

and AL are identical. Then we have

T fixa

T=

E{Rr}E{Ra}+ E{Rr} ,

T fixr

T=

E{Ra}E{Ra}+ E{Rr} . (3.38)

If we use (3.38) in (3.37), then the instantaneous e2e data rate that the dual-hop

relaying system is able to support is given by

Rfixe2e = min

{Ra E{Rr}

E{Ra}+ E{Rr} ,Rr E{Ra}

E{Ra}+ E{Rr}}. (3.39)

As a second scheme we apply Resource Optimal RA scheme. We recall from

section 3.3.2 that the resource optimal RA scheme allocates resources on

short-term basis such that Ra Ta = Rr Tr is valid in each TTI. Then we have

T opta

Ttot=

Rr

Ra +Rr,

T optr

Ttot=

Ra

Ra +Rr(3.40)

and there holds

Ropte2e =

RaRr

Ra +Rr. (3.41)

These formulas are valid for UL and DL.

Let us next consider the following formulation of the instantaneous received

SNR:

γa = γa|ha|2, γa =(Ptx,a/La)

PN,a, (3.42)

where |ha|2 refers to the channel power gain which is exponentially dis-

tributed with unitary mean. Moreover, Ptx,a, PN,a, and La denote the trans-

mission power, noise power, and mean path loss attenuation, respectively. If

a strong LoS condition is assumed in the RL (Ricean fading with large K-

factor), then the instantaneous received SNR remains close to its mean value

and approximately there holds γr =(Ptx,r/Lr)

PN,r.

We recall that for the data rates on AL and RL we use formulas

ra = Wa log2(1 + γa), Rr = Wr log2(1 + γr). (3.43)

It is noted that the macro eNB use orthogonal resources in RL to communi-

cate with the RNs in its coverage area. This condition does not hold for AL,

where the whole communication bandwidth can be re-utilized by each RN to

provide services to its associated UEs. We also note that Wa is affected by the

parameters of the open-loop power control mechanism in UL. In practice, the

value that AL bandwidth Wa may take for a given UE depends on the (max-

imum) transmission power p(ul)max, the UE-eNB distance deNB,ue, and the target

received signal power per PRB p(ul)0 which needs to be guaranteed [146].

50

Resource Optimal Relaying

Fixed (long-term) RA scheme

If time slots allocated to RL and AL, are defined by (3.38), then the mean e2e

data rate is given as

Rfixe2e = E

{Rfix

e2e

}=

Rr

Rr + E{ra}E{min

(ra,E{ra}

)}, (3.44)

where the mean data rate for the AL obtains expression

E{ra}=∫ ∞

0Wa log2(1 + γ)

e−γ/γa

γadγ=

Wa

loge(2)e1/γaE1(1/γa). (3.45)

Let us define the SNR threshold γth by

γth = exp(e1/γa E1(1/γa)

)− 1. (3.46)

This definition makes the instantaneous data rate of AL shown in (3.43) be

equal to its mean value given in (3.45). Then, we find that

E

{min

(ra,E{ra}

)}=

∫ γth

0Wa log2(1 + γ)

e−γ/γa

γadγ

+Wa log2(1+γth)

∫ ∞

γth

e−γ/γa

γadγ

=Wa

log2(e)

[− e1/γaE1

(1+x

γa

)−e−x/γa loge(1+x)

]∣∣∣∣∣γth

0

+Wa log2(1 + γth) e−γth/γa . (3.47)

Finally, plugging (3.46) into (3.47), and the resulting formula into (3.44), we

obtain

Rfixe2e =

[RrWa

loge(2) e−1/γaRr +Wa E1(1/γa)

]

×[E1(1/γa)− E1

(exp (e1/γaE1(1/γa)

)γa

)]. (3.48)

It is noted that this RA scheme requires the long-term statistics of both AL

and RL CQI (i.e., γa and Rr).

Resource Optimal (Short-term) RA scheme

If the time resources allocated to RL and AL are updated on instantaneous

basis, then the mean e2e data rate is obtained from (3.41) as

Ropte2e = E

{raRr

ra +Rr

}= Rr

[1− E

{1

1 + ra/Rr

}]. (3.49)

Closed-form expression for the expectation in RHS of (3.49) does not exist.

Hence, we compute for it an upper bound expression that will be further uti-

lized in order to obtain a tight lower bound for the corresponding e2e data

rate.

51

Resource Optimal Relaying

The RHS expectation of (3.49) can be written as

E

{[1+

raRr

]−1}= E

{[1+

Wa

loge(2)Rrloge(1+γa)

]−1}(3.50)

=

∫ ∞

0

[1+

Wa

loge(2)Rrloge(1+γ)

]−1 e−γ/γa

γadγ (3.51)

=

∫ 1

0

[1+

Wa

loge(2)Rrloge

(1−γa loge(u)

)]−1du, (3.52)

where (3.51) follows after replacing the expectation operator with its corre-

sponding integral form, while (3.52) is obtained after substitution u = e−γ/γa .

Let us use the first term of the series expansion for the logarithm presented

in (4.1.27) of [215], i.e.,

loge(z)=2

[(z−1

z+1

)+

1

3

(z−1

z+1

)3

+ . . .

]�{z} > 0. (3.53)

This formula provides a good upper bound for logarithm in 0 ≤ �{z} ≤ 1.

Then, after replacing the logarithmic function in (3.52) with first term of (3.53),

we obtain

E

{[1+

raRr

]−1}≤∫ 1

0

[1+

Wa

loge(2)Rrloge

(1−γa

2(u−1)

(u+1)

)]−1du

=

∫ 1

0

[1+

Wa

loge(2)Rrloge

(u(1−2γa)+(1+2γa)

(u+ 1)

)]−1du

=

∫ 1

0

[C2 +C1 loge

(Mu+ 1

u+ 1

)]−1du. (3.54)

For C1, C2 and M, there holds

C1 =Wa

loge(2)Rr, C2 = 1 + C1 loge(1 + 2γa), (3.55)

M =(1− 2γa

)/(1 + 2γa

). (3.56)

These are constant parameters which depend on the long-term statistics and

communication resources allocated to RL and AL.

Let v = (Mu+1)/(u+1). Then, applying this substitution in (3.54) and, after

that, using the Schwarz’s inequality we find that

E

{[1+

raRr

]−1}≤∫ 1

(M+1)/2

[C2 +C1 loge(v)

]−1 (1−M)

(v−M)2dv

≤ (1−M)

C1

{∫ 1

(M+1)/2

[C2

C1+ loge(v)

]−2

dv

} 12

×{∫ 1

(M+1)/2

[1

(v −M)2

]2dv

} 12

. (3.57)

It is possible to show that the definite integrals in (3.57) attain the following

52

Resource Optimal Relaying

closed form solutions [211]:∫ 1

u

1

[α+ loge(x)]2dx = e−α

[E1

(− α− loge(u))− E1

(− α)]

−[ 1α− u

α+ loge(u)

], (3.58)∫ 1

u

1

[x− α]4dx =

1

3

[ 1

(u− α)3− 1

(1− α)3

]. (3.59)

After applying (3.58)-(3.59) in (3.57) and plugging the resulting expression

in (3.49), the following lower bound expression is obtained:

Ropte2e ≥ Rr

[1−

{ 7

12γaC1

} 12{1− (1 + 2γa)

1 + C1 loge(1 + 2γa)

+e−1/C1

C1

[E1

(− 1

C1

)−E1

(− 1

C1−loge(1+2γa)

)]} 12

]. (3.60)

Finally, using the asymptotic expansion presented in [211]

E1(z) ∼ e−z

z

{1− 1

z+

2

z2− 6

z3+ . . .

}, (3.61)

it is possible to show that when γa grows, then the approximation

Ropte2e ∼ Rr

[1− 1

C1 loge (2γa)

]γa 1 (3.62)

becomes increasingly tight. The latter formula shows that the RL data rate

always becomes a bottleneck of the dual-hop relaying system when SNR in

the AL is large.

Simulation results

The following we present results for different RA schemes assuming certain

mean SNRs in RL and AL. It is noted that the given results are valid for both

UL and DL directions.

Figure 3.10 presents the mean e2e data rate when the macro eNBs shares

10MHz bandwidth among 10 RNs. By employing even shares, each RN has

1MHz (10MHz) to communicate in the RL (AL). Results of Figure 3.10 are

computed from the derived analytic expressions.

53

Resource Optimal Relaying

0 5 10 15 20 25 301

2

3

4

5

6

7

8

Figure 3.10. Mean e2e data rate for a dual-hop relaying system (Wr = 1MHz; Wa = 10MHz).

Red lines: γr = 5dB. Green lines: γr = 15dB. Blue lines: γr = 25dB. Dashed

lines with circles: (Long-term) fixed RA. Solid lines with squares: (Short-term)

resource optimal RA. Point values (′∗′) were simulated.

As it was expected, the resource optimal RA scheme performs better that the

fixed RA scheme and performance difference grows as the mean SNR of the

RL increases. If the RL SNR is kept constant, then the data rate gain of resource

optimal RA reduces when the mean SNR of the AL grows. In this situation

RL becomes a bottleneck for the e2e data rate. It is noted that the point wise

simulation results denoted by (′∗′) are included to validate the accuracy of the

derived closed form expressions in all cases.

Figure 3.11 and Figure 3.12 show the mean e2e data rates for UL and DL,

respectively. When compared to a RN, UE usually applies less power when

sending signals in UL. Therefore, if UL open-loop power control is applied,

then the bandwidth that a UE can utilize in AL (Wa) is usually smaller than

the bandwidth that the RN uses in RL (Wr).

54

Resource Optimal Relaying

0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

Figure 3.11. Mean e2e data rate for a dual-hop relaying system (Wr = 1MHz) for Uplink di-

rection (γr = 5dB). Red lines: γa = 10dB. Green lines: γa = 20dB. Blue lines:

γa = 30dB. Dashed lines with circles: (Long-term) fixed RA. Solid lines with

squares: (Short-term) resource optimal RA.

1 2 3 4 5 6 7 8 9 101

2

3

4

5

6

7

Figure 3.12. Mean e2e data rate for a dual-hop relaying system (Wr = 1MHz) for Downlink

direction (γr = 20dB). Red lines: γa = 10dB. Green lines: γa = 20dB. Blue

lines: γa = 30dB. Dashed lines with circles: (Long-term) fixed RA. Solid lines

with squares: (Short-term) resource optimal RA.

In DL the macro eNB transmits at constant power towards multiple RNs.

Then, the ratio between Wa and Wr, is expected to be smaller (larger) than 1

for the UL (DL) communication. For the DL direction, the performance gain

55

Resource Optimal Relaying

of resource optimal RA over fixed RA scheme vanishes when Wa/Wr → 1.

However, in the UL direction, the performance is maximized in this situation.

Hence, it can be concluded that the AL represents the bottleneck in the UL

communication, whereas, RL is the limiting factor in the DL communication.

3.5 Conclusions

We analyzed resource allocation schemes (i.e., conventional RA, fixed RA with

and without buffer, and the resource optimal RA) in a two-hop relaying. It

was observed that the resource optimal RA provides notable performance

gains when compared to other RA schemes, where radio resources are not

allocated in short-term. The main contributions of this chapter include the

derivation of closed-form expressions of the CDF and PDF of the e2e data

rate, as well as expressions for the average and outage data rates. Moreover,

an accurate lower bound for the mean data rate was also deduced.

The derivation of closed-form expressions for e2e data rates was repeated for

the DL/UL decoupled scenario to study the obtained e2e data rates when the

fixed (long-term) and resource optimal (short-term) RA schemes are applied

in UL and DL directions. The e2e data rate depends not only on the SNR

that both RL and AL experience, but it also depends on the amount of radio

resources that are allocated to each link. If the UL and DL decoupling is em-

ployed, then, the derived formulas determine the most optimum UE to RN

association (and RN association to given DeNB) for the communications in

each direction.

Overall, it was observed that the resource optimal RA scheme can have signif-

icant performance enhancement due to the imbalance in resourcing between

RL and AL. The use of resource optimal RA schemes improve the RL effi-

ciency, which indirectly leads to a positive impact on the e2e data rate perfor-

mance. The obtained results shed light into the complex relationship between

RL and AL design parameters and system variables, and hence, they poten-

tially support the dimensioning of the infrastructure relaying system.

56

4. Practical Interference Mitigation forthe Relay Backhaul Link

4.1 Introduction

The LTE-A Relay Node employs self-backhauling towards the serving donor

eNode B (DeNB) via the Un radio interface making relaying an attractive so-

lution for the operators [216]. However, Relay Link may also represent a ca-

pacity bottleneck [20]. In LTE-A, the inband relaying utilizes a time-based

resource scheduling between RL and AL. In addition, RN competes on the

radio resources with UEs directly served by the DeNB via a direct link. More-

over, the competition for radio resources further increases when the DeNB is

serving more than one RN. As a result, scarce radio resources are heavily com-

peted and RL may become congested. We also note that RL is experiencing

interference from the neighbouring eNBs, especially when the RN is deployed

at the cell edge. At the same time all RNs can serve their UEs by utilizing the

same radio resource pool in the AL. Then, the spatial radio reuse is much more

effective for the AL resources than it is for the RL resources. The resulting im-

balance between RL and AL can be compensated e.g. by advanced antenna

technologies, coordinated communication and/or by carefully planning the

relay locations [19,24,217–219]. These methods are effective but will increase

the cost and implementation complexity of the relaying.

As mentioned, there is interference from adjacent eNBs on the RL. In some

locations RL may even experience a Line of Sight (LoS) or a dominant signal

direction towards the interfering eNB. In these cases the interference easily

becomes crucial [37, 220, 221]. To that end, this chapter includes a study of a

practical method that can be used on RL to improve the desired signal strength

from DeNB and/or to mitigate the interference from the interfering eNB. We

present an approach where a few bit channel feedback from the RN to both the

DeNB and the interfering eNBs is employed. We note that a simple version of

57

Practical Interference Mitigation for the Relay Backhaul Link

the proposed feedback method has been already standardized in 3G and 4G.

Obtained results has been summarized in our previous publications [12, 13].

The rest of chapter is organized as follows. A brief literature survey is pro-

vided in Section 4.2. Section 4.3 presents the considered system model to ad-

dress the RL limitations. Moreover, Section 4.4 provides the analysis of the

SINR and e2e outage rate obtained on RL. Furthermore, mathematical equa-

tions are validated via simulations in Section 4.5. Work is summarized in

Section 4.6.

4.2 Previous Work and Contributions

Several studies have proposed various improvements for the RL performance.

To that end, Bulakci et al. [18] and Saleh et al. in [19] proposed network plan-

ning techniques namely location selection and cell selection for the improved

RL performance. The former technique enables RNs to be deployed in less

shadowed location while in the latter the RN is served by the best available

eNB. However, the good location options for the RN deployment are limited

especially in dense urban areas due to high building obstruction. In [30],

Park et al. evaluated the relaying performance by employing a dual-layer

BF (DLBF) on RL assuming that the resources allocation is optimized. In ad-

dition, higher order MIMO was employed on the AL of spatially multiplexed

RNs. Results show that the proposed scheme enables significant system gains

of upto 241% as compared to 2×2 MIMO in the RL. However, this work only

considers the transmission of DeNB towards the serving the RNs by employ-

ing MIMO antennas, while not addressing the interference from the neigh-

bouring eNBs.

Yi et al. in [25] present a two-step resource allocation scheme to enhance the

RL performance. The proposed resource assignment model take into account

the link quality of RL, AL and macro UEs. Result shows that the proposed

scheme enhance the average cell performance and the performance of the cell-

edge UE. In [26], a downlink resource scheduling scheme was proposed for

the TD-LTE relaying network. The proposed scheme allocate dynamically the

frequency resources on direct and AL on need basis. The resources for RL are

allocated using the user queue information.

Furthermore, Li et al. in [31] examine the utilization of virtual MIMO in or-

der to enhance the RL performance in dense relaying networks. This work

employs the MIMO technique on cluster of RNs such that all RNs form a vir-

tual MIMO on RL towards the base station. This work mainly focus on the

58

Practical Interference Mitigation for the Relay Backhaul Link

out-band relaying where the DeNB and RN operate on exclusive frequency

bandwidths. This not only leads to spectrum scarcity due to exclusive allo-

cation of spectrum for RL but also increase the system complexity and con-

trol signaling among the RNs in the cluster. Luo et al. in [32] proposed an

iterative Co-Phasing Water-filling (ICPWF) scheme in order to enable the op-

timal power solutions in order to improve the RL performance. Here, two

base stations simultaneously carry out coordinated transmissions to a RN

deployed at the cell edge. Moreover, Haile et al. in [34, 35] employs the

coordinated multi-point transmission in order to relax the self-backhauling

bottlenecks of RL. Results show improved system performance in terms of

e2e rates. However, the method induce additional control signaling as well

as increase system complexity by engaging several eNBs. In addition to

the mentioned works, there is a large number of studies in the field, see

e.g. [131, 184, 187, 188, 192, 199, 222, 223]. Yet, to the best of our knowledge,

our simple, practical and standard compliant method has not been studied by

other authors.

The contributions of this chapter are the following: We derive analytical ex-

pressions for the RL SINR distributions assuming Rice and Rayleigh fading

combinations on the dedicated and interfering links, and a certain interfer-

ence mitigation method based on limited feedback. In addition to SINR dis-

tributions, we also deduce the e2e outage probabilities for the assumed in-

terference and feedback cases. Analytical results are verified via numerical

simulations.

4.3 System Model

We focus on the downlink of a REC network with a dual-hop relaying system

as shown in Figure 4.1.

4.3.1 System Assumptions and Definitions

Fast and Slow Feedback

If the Channel Quality Information (CQI) feedback, sent by the receiver, ar-

rives to the transmitter and is applied within the duration of the channel co-

herence time, then feedback is said to be ’fast’. It is noted that the channel

coherence time is inversely related to the transmitter/receiver motion. In con-

trast to fast feedback, if the CQI feedback requires clearly more time than the

channel coherence time, then it is said to be ’slow’. This long-term feedback

59

Practical Interference Mitigation for the Relay Backhaul Link

Serving NeighborRelaynode

Rice/Rayleigh block fading channel

UE

Rr , Wr , d i

Ra , Wa , a

Fast/slow feedback Fast/slow feedback

Strongest interferer

Irest

DeNB eNB

Other eNBs

Figure 4.1. System model.

may describe e.g. channel mean, variance, or correlations between component

channels. To obtain a reliable long-term information about the channel, some

filtering is usually needed. Furthermore, this filtering should be used over

several channel coherence times in order to mitigate the impact of short-term

channel variations.

Channel Models

Unless otherwise stated, we assume that RN and eNB locations are fixed. To

that end eNB antennas are elevated above the street level where the surround-

ing environment around the eNB remains unchanged. RN is located closer to

the street level but it may experience a LoS towards the DeNB due to e.g., loca-

tion planning. Yet, the moving objects around the RN within the first Fresnel

zone may create a fast fading on RL. Then, the Rice fading provides a suitable

option for modeling the radio channel.

If there is no LoS on RL, then RN receives large part of signal power from

several directions rather than from a single direction. Such a channel envi-

ronment can be described using the Rayleigh channel model. Other options

might be Nakagami-m model [224], [225] or the so-called general fading mod-

els like Stacy model [226], [227]. In case of these general channel models a

problem arise, when values for numerous parameters are chosen to define the

fading. Thus, we have chosen an approach where results are deduced for the

Rice and Rayleigh models. We note that these models provide a reasonable

basis for the link and system performance analysis.

In case of Rayleigh model, the channel components hm of the vector h are i.i.d.

complex zero-mean Gaussian random variables, meaning that the |hm| fol-

lows the Rayleigh distribution. The channel feedback for the case of Rayleigh

model will be fast. In Rician model, the channel components of the vector h

60

Practical Interference Mitigation for the Relay Backhaul Link

are i.i.d. complex Gaussian with the same mean |E{hm}| = ν ∈ R+, which

means that the |hm| follows the Rice distribution. In the following, we use

normalized channels so that E{|hm|2} = 2σ2+ν2 = 1, where 2σ2 is the power

of the fading component and ν2 is the power of the static signal component.

Here, the RN will employ the slow feedback approach on RL to provide CQI

to eNB.

System Assumptions

1) We assume the block fading model on both RL and AL.

2) We employ the AWGN channel capacity formula in order to compute the

instantaneous rates on the RL (Rr) and the AL (Ra):

Rr = Wr · log2(1 + Υr), Ra = Wa · log2(1 + Υa), (4.1)

where Wr and Wa are the transmission bandwidths and Υr and Υa are the

SINR’s on RL and AL, respectively. We need the distributions of Υr and Υa in

order to analyze the e2e performance.

3) It is noted that the experienced SINR’s Υr and Υa on RL and AL, respec-

tively, are independent and hence, the experienced data rates Rr and Ra on

corresponding links are also independent.

4.3.2 End-to-end outage rate

We recall the conventional e2e rate (Re2e) formula to initiate the mathematical

treatment of the relaying system;

Re2e =1

2min{Rr, Ra}. (4.2)

To that end, we compute the outage rate to examine the e2e rate for a relaying

system;

Pout = P (Re2e < Rmin) = FRe2e(Rmin). (4.3)

Here Rmin is the minimum rate requirement that can be used to control the

e2e performance. That is, link is considered to be in outage if Re2e < Rmin.

Evidently, the outage probability Pout for such an event should be small. To

compute the outage rate we need the CDF of the e2e rate, denoted in (4.3) by

FRe2e . While assuming that instantaneous rates in RL and AL are independent

random variables, we have

FRe2e(R) = 1− P (min{Rr, Ra} > 2R)

= 1− P (Rr > 2R)P (Ra > 2R)

= 1− (1− FRr(2R))(1− FRa(2R)),

(4.4)

61

Practical Interference Mitigation for the Relay Backhaul Link

where FRr and FRa denote the CDF’s of RL and AL rates, respectively. Thus,

we need distributions of the Rr and Ra in order to deduce the e2e rate distri-

bution. To that end, the expression in (4.1) can be used to obtain the RL rate

distribution as;

FRr(R) = P (Rr < R) = P (Wr · log2(1 + Υr) < R)

= P (Υr < 2R/Wr − 1) = FΥr(2R/Wr − 1).

(4.5)

In the same way, we can also deduce the AL distribution as FRa(R) =

FΥa(2R/Wa − 1). Hence, in order to compute the outage rate, we need the

SINR CDF’s of RL and AL.

It is noted that the fixed RN deployment is usually employed at the cell edge to

improve the cell coverage [6,67]. Then, the macro DeNB will obviously create

co-channel interference to a large area and consume scarce time/frequency

radio resources for RL transmission. While there is need to minimize the uti-

lization of radio resources on RL, it may easily make RL as a bottleneck in the

e2e performance especially if the network is interference limited. In such sce-

nario each RN will provide channel feedback to its interfering eNB in order to

mitigate the interference from the interfering eNB through precoding. To that

end, it is assumed that the serving and the interfering eNBs are synchronized

and utilize the same time frequency resource pool for RN transmissions.

The following sections will present the considered precoding methods. These

methods will be used to mitigate the interference from dominant interfering

eNB and to improve the RL quality from the dedicated DeNB.

4.3.3 RL SINR Model

Let us consider the SINR in the RN as;

Υn,k =γn,n,k|wn · hn,n,k|2

1 +∑

m�=n γm,n,k|wm · hm,n,k|2 . (4.6)

Here indices m and n refer to eNBs in mth and nth cells, respectively, while k

refers to the RN of the nth cell and wm is the normalized precoding weight

vector used in the cell m. It is noted that hm,n,k represents the normalized

vector channel between mth eNB and kth RN of the nth cell. We also note that

the dimensions of the weight and channel vectors are defined by the number

of parallel transmission chains in the eNB. We haveE{|wn ·hn,n,k|2} = 1 due to

normalizations, if precoding vector is randomly selected. Furthermore, γm,n,k

denotes the average received SNR in the RN. The equation (4.6) can be written

as follows:

Υn,k =γn,n,k|wn · hn,n,k|2

1 + γn,n,k|wn · hn,n,k|2 + Irestn,k

, (4.7)

62

Practical Interference Mitigation for the Relay Backhaul Link

where γn,n,k = maxm�=n{γm,n,k} denotes the mean power of the

strongest/dominant interfering signal and Irestn,k includes the rest of the

interfering signals,

Irestn,k =

∑m�=n,n

γm,n,k|wm · hm,n,k|2. (4.8)

As can be seen from formulae (4.7)–(4.8), the eNB multiantenna resources can

employed as follows:

1) We can enhance signal strength transmitted from the DeNB towards its own

RN by selecting the precoding weight such that |wn ·hn,n,k|2 is maximized, i.e.

|wn · hn,n,k|2 = max{|w · hn,n,k|2 : w ∈ W}, (4.9)

whereas W denotes the precoding codebook. This is a baseline scenario where

RN provides CQI feedback to the DeNB only.

2) If there is a dominant interfering eNB such that γn,n,k >> γm,n,k, m �= n,

then RN SINR performance will be enhanced if the dominant interfering eNB

n employs its antenna resources to mitigate the interference towards RN as

γn,n,k|wn ·hn,n,k|2 in (4.7) by e.g. selecting the precoding weight from the code-

book such that

|wn · hn,n,k|2 = min{|w · hn,n,k|2 : w ∈ W}. (4.10)

It is obvious that such action will diminish the BF gain in RL between the

dominant interfering eNB n and its own RN. Though, this may not be nec-

essarily a problem if the RL between the dominant interfering eNB n and its

own RN is good or precoding weight that maximize the own RL and minimize

the interference is the same. Furthermore, if there is a strong LoS component

on RL, then there is some directivity by nature and due to fixed location of

RN, the LoS component is quite static. Hence, in this scenario, the RL radio

resources of different eNBs can be jointly scheduled in order to minimize the

interference towards adjacent cells RN.

In the following we study the performance when eNB applies the so-called al-

truistic precoding to mitigate the interference towards the RL of the adjacent

macro eNBs. The altruistic transmit BF has been previously investigated

in [228, 229], though, those studies mainly focus on femtocells in NLoS sce-

nario with Rayleigh fading channel. We fit some results of [228, 229] in our

context for comparison purposes, but our study mainly focus on Rice fading

experienced by a fixed RN on the RL.

63

Practical Interference Mitigation for the Relay Backhaul Link

Before considering the precoding in more details we write (4.7) in the form

Υn,k =γn,n,k|wn · hn,n,k|2

1 + γn,n,k|wn · hn,n,k|2 + Irestn,k

=

γn,n,k1+Irest

n,k|wn · hn,n,k|2

1 +γn,n,k1+Irest

n,k|wn · hn,n,k|2

(1)n,kγn,n,k|wn · hn,n,k|2

1 + Ψ(1)n,kγn,n,k|wn · hn,n,k|2

,

(4.11)

where factor Ψ(1)n,k = 1/(1 + Irest

n,k ) reflects the structure of the interference re-

ceived in kth RN of nth macrocell. Here the superscript ’1’ indicates that the

strongest interfering signal has not been taken into account. We note that

0 < Ψ(1)n,k < 1 and the closer Ψ

(1)n,k is to one, the more dominant the strongest

interferer is. If interference mitigation is applied only for the strongest inter-

fering signal, then we may approximate

Irestn,k ≈ E{Irest

n,k } =∑

m�=n,n

γm,n,kE{|wm · hm,n,k|2}

=∑

m�=n,n

γm,n,k = Irestn,k ,

Ψ(1)n,k ≈ 1/(1 + Irest

n,k ),

(4.12)

whereas E{·} denotes the expectation over fast fading. This approximation is

most valid when the strongest interfering eNB is clearly dominant.

We are left with two random variables after approximation (4.12), including

the gain factor for the desired signal from DeNB: |wn · hn,n,k|2; and the gain

factor for the interfering signal from dominant interfering eNB: |wn · hn,n,k|2.We analyze the maximum achievable performance gain through use of some

simple and practical precoding methods by applying (4.9) and (4.10) when se-

lecting precoding weights for the desired DeNB and the dominant interfering

eNB. Now, we re-write SINR (4.6) in the form

Υr ≈ Ψ(1)γd|w · hd|21 + Ψ(1)γi|w · hi|2

, (4.13)

where equality has been replaced by approximation due to (4.12), subscript

′d′ refers to desired signal and subscript ′i′ refers to the interfering signal. The

SINR model (4.13) provides our starting point for the analytical study in Sec-

tion 4.4.

Fast and Slow Quantized Co-Phasing (QCP)

Before further mathematical analysis, we discuss briefly about the applied

precoding schemes. In the previous literature, the use of precoding for the

64

Practical Interference Mitigation for the Relay Backhaul Link

purpose of interference mitigation has been examined, e.g., in [229–231].

In the current study, we assume a simple method with two antennas, see

also [232]. While approach is simplistic, it is very practical and tractable math-

ematical analysis can be used to provide insight to the link level performance.

The applied precoding methods are known as Fast QCP (FQCP) and Slow

QCP (SQCP).

In FQCP scheme, we assume two-antenna transmitter and a single antenna

receiver. We also assume N bits of phase information available at the trans-

mitter. Then w = (w1, w2) = (1, ejϕ)/√2 for FQCP is defined by

|w · h| = maxn

{|(h1 + ejϕnh2)|/√2}, (4.14)

where ϕn = 2π(n− 1)/2N and n = 1, 2, . . . , 2N .

The work done in [229, 232, 233] already employed FQCP which can be ex-

tended to any number of transmit antennas. Although FQCP is simple and

widely known, its performance on non-Rayleigh channels has not been well

studied. We note that in interference mitigation we simply select w =

(w1, w2) = (1, ejϕ)/√2 such that minimum is obtained in (4.14) instead of

maximum.

We assume a single antenna receiver and two antenna transmitter with N bits

of phase information available for SQCP case. The weight w = (w1, w2) =

(1, ejϕ)/√2 for SQCP is selected as in FQCP but it is based on long-term mean

of the channel:

|w · h| = maxn

{|(E{h1}+ ejϕnE{h2})|/√2}, (4.15)

where ϕn = 2π(n− 1)/2N and n = 1, 2, . . . , 2N . In case of the Ricean channel

model E{h1} = ν1ejψ1 , E{h2} = ν2e

jψ2 and (4.15) becomes

|w · h| = maxn

{|ν1 + ej(ϕn−ψ1+ψ2)ν2|/√2}. (4.16)

If transmitter antennas are grouped in the same mast, then ν1 = ν2 =: ν. Sim-

ilarly as in FQCP, we select w = (w1, w2) = (1, ejϕ)/√2 such that minimum is

obtained in (4.15) instead of maximum for interference mitigation. The main

aim of this interference mitigation is to mitigate the Rice component of the

interfering signal. Moreover, the slow CQI feedback update does not impose

capacity limitations. Finally, we note that the phasing accuracy is not neces-

sarily defined only by N , but also by the accuracy of the phase measurements

in the receiver.

65

Practical Interference Mitigation for the Relay Backhaul Link

4.4 Analysis of SINR and e2e Outage Rate

We can write the SINR Υr of RL in the form of X/(1 + Y ) according to (4.13),

whereas random variables X = Ψ(1)γd|w · hd|2 and Y = Ψ(1)γi|w · hi|2 are

independent. Here γd and γi denote the mean powers of the dedicated RL

channel from desired DeNB and the interfering RL channel from interfering

eNB, respectively, before feedback is used to adjust the channel components.

By [212], we have

FΥr(γ) =

∫ ∞

1FX(γt) fY (t− 1)dt, (4.17)

where fY (y) is the PDF of Y and FX(x) represents the CDF of X . The chal-

lenge here is to find the distributions forX and Y , and then compute the SINR

distribution from (4.17). If there is no interference, then we can write

FΥr(γ) = Fγd|w·hd|2(γ). (4.18)

In the following F cΥr(γ) = 1− FΥr(γ) is the Complementary Cumulative Dis-

tribution Function (CCDF). We consider four scenarios with different fading

model combinations in the dedicated RL and the interfering RL in the coming

analysis. In each fading scenario the channel model for the dedicated RL will

be mentioned first, while the model for the dominant interfering eNB is given

second.

4.4.1 Scenario 1: Rice fading - Rice fading

Relay system framework

If RN experiences a LoS on dedicated RL and in the interfering link, then the

Rice fading model is a suitable option for both to demonstrate the channel

statistics. Although the RN location is fixed, some fading can occur due to

the movement of scatterers (e.g., vehicles, etc) nearby the RN. This scenario

assumes slow feedback scheme (i.e., SQCP) on either/both dedicated RL and

the interfering link.

SINR distribution in the presence of SQCP

In the Rice fading modeling, the component channel related to each transmit

antenna m is of the form hm = αm + νejψm , where αm is complex zero-mean

Gaussian which means that |αm| follows the Rayleigh distribution. The sec-

ond term in the sum (i.e., νejψm) denotes the static channel component, i.e.

ν > 0 is constant and ν2 is the power of the LoS part of the signal. Angles

ψm are assumed to be mutually independent and uniformly distributed on

(−π, π), while it is noted that ψm in each antenna branch can take any value

66

Practical Interference Mitigation for the Relay Backhaul Link

on (−π, π) if antenna branches are not calibrated - as usually is the case in

practice. Yet, the value of ψm is changing only slowly - if at all. We recall that

the K = ν2/2σ2, 2σ2 = E{|αm|2} denotes the so-called Rice factor (K), which

is the ratio between the power of the fixed-path (i.e., ν2) and the power of the

fading part (i.e., 2σ2) of the signal. In the following, the component channels

of h (whether it is desired or interfering vector channel) are normalized such

that

E{|h1|2} = E{|h2|2} = 2σ2 + ν2 = 1 (4.19)

and the Rice factors for the component channels are equal: K = ν2/2σ2. Then,

we can write the powers of fading and static channel parts as

2σ2 =1

1 +Kν2 =

K

1 +K. (4.20)

Let us consider the distribution of |w · hd| when SQCP is applied on the ded-

icated RL. More precisely, we will show the following: If h1 and h2 are i.i.d.

and defined as above, and w is selected using the SQCP method in (4.15), then

|w · hd| follows the Rice distribution with parameters

ν := |E{w · hd}| ≈√

1 + sinc( 1

2Nd

) · νd,K :=

|E{w · hd}|2E{|w · hd − E{w · hd}|2} ≈

(1 + sinc

( 1

2Nd

))·Kd.

(4.21)

Hence, the sum channel will be Rician after using SQCP as defined in (4.15)

over two Rician channels. Furthermore, the parameters ν and K of the sum

channel can be expressed in terms of the component channel parameters νd,

Kd and the number Nd of phase bits used in SQCP.

To prove this result, we assume that w is selected according to (4.15). Then

we can write

w·hd =1√2(α1 + νde

jψ1) +ejϕ√2(α2 + νde

jψ2)

=( 1√

2α1 +

1√2ejϕα2

)+

νd√2

(ejψ1 + ej(ϕ+ψ2)

),

(4.22)

where the first term in the right side is a complex zero-mean Gaussian since

both α1 and α2 are complex zero-mean Gaussian and the phasing is using

only long-term channel information. Moreover, the second sum on the right

side is static - from fast fading perspective - and we find that |w · hd| follows

the Rice distribution. We also note that E{|(α1 + ejϕα2)/√2|2} = 2σ2

d, which

means that the power of the fading part of the signal remains the same after

using the SQCP.

Furthermore, we need to prove (4.21). We take an expectation with respect to

the fast fading over (4.22). Since1√2(α1 + ejϕα2) is a zero-mean variable, we

67

Practical Interference Mitigation for the Relay Backhaul Link

obtain

|E{w · hd}|2 = ν2d2

∣∣∣ejψ1 + ej(ϕ+ψ2)∣∣∣2

=ν2d2|1 + ejΦ|2

= ν2d · (1 + cosΦ),

(4.23)

whereas we haveΦ = ϕ+ψ2−ψ1. Althoughψ1 andψ2 change slowly, they are

random due to non-ideal practical implementation. For example, phase drift

is a well-known phenomenon in practical transmission chains. By applying

the SQCP, we select ϕ based on Nd feedback bits such that Φ = ϕ + ψ2 − ψ1

maximizes the value of cosΦ. Due to quantization of ϕ the phase Φ becomes

uniformly distributed in (−π/2Nd , π/2Nd). Now we can write

|E{w · hd}|2 = ν2d · (1 + EΦ{cosΦ}) + Δ, (4.24)

where Δ = ν2d ·(cosΦ−EΦ{cosΦ}) and EΦ refers to the long term expectation

over Φ. After simple calculations, we find that EΦ{cosΦ} = sinc(1/2Nd). In

case of SQCP, accurate co-phasing is well possible because phasing between

static signal components varies on slowly basis. Then Δ becomes small and

first approximation in (4.21) is valid. To show the second part of (4.21), we

compute

E{|w · hd−E{w · hd}|2}= E{|(α1 + ejϕα2)/

√2|2} = 2σ2

d

(4.25)

and we have

K =ν2

2σ2d

≈(1 + sinc

(1

2Nd

))ν2d

2σ2d

=(1 + sinc

( 1

2Nd

))·Kd.

(4.26)

Hence, it is proved that after applying SQCP, the received signal amplitude,

i.e., |w · hd| follows the Rice distribution with parameters (4.21). The CDF of

the Rice variable with parameters ν and σd is by [207] of the form

Fγd|w·hd|(γ) = 1−Q1

σd,

√γ

γdσ2d

), (4.27)

where Q1 is the well-known Marcum Q-function defined as

Q1(A,B) =

∫ ∞

Bte−

t2+A2

2 I0(At)dt (4.28)

and I0 refers to the modified Bessel function of order zero [211]. By (4.20),

(4.21) and (4.27), we obtain the CCDF

F cΥr

(γ) = Q1

(√2(1 + sinc

( 1

2Nd

))Kd,

√2γ

γd(1 +Kd)

). (4.29)

68

Practical Interference Mitigation for the Relay Backhaul Link

SINR distribution when SQCP is applied both in the desired RL and in the

interfering link

Here, we discuss the scenario where the interfering eNB is in LoS with the

RN. Though, we may use the SQCP scheme for mitigating the interference

from the interfering eNB. Then, similarly with (4.21) in the previous section

we can deduce the Rice factor of the interfering channel

K =ν2

2σ2i

=(1− sinc

( 1

2Ni

))·Ki, (4.30)

where the 2σ2i andKi denote the power of the fading part and the Rice factor of

the component channels. We obtain K/Ki = 0.026 with just three bit phasing,

which means that with three bit phasing accuracy we may attenuate 16 dB the

Rice factor of the signal from interfering eNB.

Since the Rice factor can be heavily attenuated just with few bit phasing accu-

racy, we can ignore it in the following and assume that after applying SQCP,

the received interference channel from interfering eNB is zero-mean Gaus-

sian. That is, in (4.13) we now haveY = Ψ(1)γi|w·hi|2, whereE{|w·hi|2} ≈ 2σ2i

and we approximate

fY (y) ≈ 1

Ψ(1)γi2σ2i

e−y/Ψ(1)γi2σ2i . (4.31)

Let us compute the CCDF of SINR Υr. Using (4.17), (4.37), (4.27) and (4.31),

we find that

F cΥr(γ) =

e1

Ψ(1)γi2σ2i

Ψ(1)γi2σ2i

∫ ∞

1Q1

σd,

√γt

Ψ(1)γdσ2d

)e− t

Ψ(1)γi2σ2i dt. (4.32)

Analytic computation of the integral in (4.32) is challenging but fortunately

in [234] a new formula has been deduced for the Marcum Q-function that

enables series presentation for the integral. According to [234] there holds

Qm(A,B) =

(A2

2

)1−m

e−A2+B2

2 Φ3

(1, 2−m;

A2

2,A2B2

4

), (4.33)

where Φ3 is the confluent hypergeometric function of two variables [211]

(9.261.3):

Φ3

(C,D;w, z

)=

∞∑m,n=0

(C)m(D)m+n

zmwn

m!n!. (4.34)

Here (C)m = Γ(m+C)/Γ(C) is the Pochhammer symbol. To compute F cΥr(γ)

we combine (4.33), (4.34), and apply standard integration rules. The resulting

expression is of the form

F cΥ(γ) =

2c

b2e−

a2+b2

2

∞∑m,n=0

1

(m+ n)!

(a22

)m+n en

(b2

2 + c)

(1 + 2c

b2

)n+1 , (4.35)

69

Practical Interference Mitigation for the Relay Backhaul Link

where en(z) =∑n

k=0 zk/k! and

a2 =ν2

σ2d

= 2(1 + sinc

( 1

2Nd

))Kd,

b2 =γ

σ2dΨ

(1)γd=

2γ(1 +Kd)

Ψ(1)γd,

c =1

Ψ(1)γi2σ2i

=1 +Ki

Ψ(1)γi.

(4.36)

4.4.2 Scenario 2: Rayleigh fading - Rayleigh fading

Relay system framework

Generally, the Rayleigh fading model is suitable channel statistics for the RL of

a mobile RN. Though, if RN admit fixed location then we can use the Rayleigh

statistics, if RN is experiencing a NLoS link towards the eNB and there are

moving scatterers in close vicinity of the RN. To that end, we may apply FQCP

in either/both the dedicated RL and the interfering link. Here, we can apply

previous results of [229, 232], where a direct connection between the BS and

the UE was assumed. This scenario will be used as a benchmark in perfor-

mance studies of Section 4.5.

SINR distribution in the presence of FQCP

Now we assume that the components of the vector channels hd and hi are

complex zero-mean Gaussian. Then, an effective chi-square approximations

for X and Y can be used according to [229] as follows;

FX(x) = 1−(1 +

2x

EΨ(1)γd

)e−2x/EΨ(1)γd ,

fY (y) =1

EΨ(1)γie−y/EΨ(1)γi ,

(4.37)

where E := E{|w · hd|2} and E := E{|w · hi|2}. The resulting expressions are

E = 1 +π

4· sinc

( 1

2Nd

), E = 1− π

4· sinc

( 1

2Ni

), (4.38)

where Nd and Ni denotes the numbers of phase bits used for the construc-

tive and the destructive phasing in the dedicated and the interfering links,

respectively.

The distribution of SINR in (4.13) can be computed using (4.17) and approxi-

mations (4.37). Applying the results of [229] we obtain CCDF of the form

F cΥr(γ) =

(2γ · ρ(ρ+2γ

)2 +

(1+ 2γ

EΨ(1)γd

ρ+2γ

)e

−2γ

EΨ(1)γd , (4.39)

where

ρ =γdγi

· 1 +π4 · sinc

(1/2Nd

)1− π

4 · sinc(1/2Ni

) . (4.40)

70

Practical Interference Mitigation for the Relay Backhaul Link

4.4.3 Scenario 3: Rice fading - Rayleigh fading

Relay system framework

In REC network planning, the RN location can be selected with aim to expe-

rience a good radio signal towards the DeNB on RL as well as to avoid inter-

ference from the adjacent eNB [19, 24]. Due to the said network planning, a

scenario is typical where RN admit the LoS Rice channel towards the serving

DeNB and NLoS Rayleigh channel towards the (dominant) interfering eNB.

Then, in addition to SQCP in the dedicated RL, the use of FQCP can be con-

sidered in the interfering link.

SINR distribution when SQCP is applied in the desired RL and FQCP is applied in

the interfering link

In this scenario we assume that the interference channel between the inter-

fering eNB and RN is complex zero-mean Gaussian. Now, if a fast feedback

channel between RN and the interfering eNB exists, then FQCP can be ap-

plied for interference mitigation. Thus, we can apply chi-square approxima-

tion as in (4.37) with E = 1− π4 · sinc(2−Ni). On the other hand, if there is no

fast feedback channel, then the interfering signal component channels sum up

randomly and the amplitude of the interference channel follows the standard

Rayleigh distribution, i.e. E = 1 in the latter formula of (4.37).

Here, the CCDF of SINR is obtained from (4.35). Since SQCP is used in the

desired RL the parameters a and b are as in (4.36). Furthermore, since FQCP

is applied in the interfering channel, the parameter c becomes

c =1

Ψ(1)γiE=

1

Ψ(1)γi(1− π

4 · sinc(

12Ni

)) . (4.41)

4.4.4 Scenario 4: Rayleigh fading - Rice fading

Relay system framework

Assume a mobile RN (located e.g. in a bus), where RL between DeNB and

RN can be for a while blocked by e.g., building, while simultaneously a LoS

towards the interfering eNB takes place. It is obvious that such a situation

usually leads to a handover, but it is noted that handover process requires

a time window during which the RN is required to operate in unfavorable

channel conditions. It is not expected that this scenario is common in the

network.

71

Practical Interference Mitigation for the Relay Backhaul Link

SINR distribution when FQCP is applied in the desired RL and SQCP is applied in

the interfering link

Here, the SNR distribution for the desired RL is given by the first formula

in (4.37). After using SQCP, we assume as before that the interference (sum)

channel is complex zero-mean Gaussian and the PDF (4.31) can be applied.

Similarly as in Section 4.4.2, we obtain

F cΥr(γ) =

(2γ · ρ(ρ+2γ

)2 +

(1+ 2γ

EΨ(1)γd

ρ+2γ

)e

−2γ

EΨ(1)γd , (4.42)

where E = 1 + π/4 · sinc(1/2Nd) by (4.38) and

ρ =γdγi

(1 +Ki

)(1 +

π

4· sinc

( 1

2Nd

)). (4.43)

4.4.5 End-to-end outage rate

As it was discussed in Section 4.3.2, we need the RL and AL CDF functions

FΥr(γ) and FΥa(γ) for the e2e outage analysis. More precisely, by (4.4) and

(4.5) we have

FRe2e(R) = 1− (1− FRr(2R))(1− FRa(2R))

= 1− F cΥr(22R/Wr − 1) · F c

Υa(22R/Wa − 1).

(4.44)

In previous sections we obtained the FΥr(γ) for different channel, interfer-

ence and feedback scenarios. Since our focus is on the impact of the feedback

and interference on RL, we consider the AL as interference free link. Then

in Rayleigh and Rice fading cases the CDF’s of the AL are respectively of the

form

FΥa(γ) = 1− eγ/γa ,

FΥa(γ) = 1−Q1

(νaσa

,

√γ

γaσ2a

),

(4.45)

where γa is the mean AL SNR and νa, σa refer to the AL Rice parameters.

4.5 Performance Evaluation

This section presents the performance evaluation of the aforementioned sce-

narios. The results include the CDF plots of SINR and e2e outage rate ob-

tained in Section 4.4. All curves are computed using previously deduced

mathematical functions. Table 4.1 summarizes the parameters used in sim-

ulations.

72

Practical Interference Mitigation for the Relay Backhaul Link

Table 4.1. Parameters for the performance experiments.

Parameter γd γi Ψ(1) Nd, Ni Kd, Ki

Value 12 dB 6 dB 0 dB/-7 dB 4 bits 9 dB

It is noted thatΨ(1) becomes 0 dB if Irest = 0, which means that RN experience

interference on RL only from the dominant interfering eNB. In contrast, if the

average powers of the dominant interfering eNB and the rest of the interfer-

ence are equal, then Irest=6 dB and we have Ψ(1) = -7 dB. Moreover, it is also

noted that figures include point values (presented in figures by ’o’, ’�’ and

’�’) which has been simulated. Simulated results are well consistent with the

curves obtained using the equations.

4.5.1 RL SINR Distributions

Scenario 1: Rice fading - Rice fading

Figure 4.2 shows the CDF of SINR on RL when we applied the SQCP and

parameters of Table 4.1. As a performance upper bound we have used the

case where SQCP is applied on desired RL and there is no interference at

all (denoted by dotted curve with ’�’). Curve was generated by using the

equation (4.29).

0 2 4 6 8 10 12 14 16 18 200

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Figure 4.2. CDF of the SINR of RL when Rice fading occurs in both the desired RL and the

interfering link, and parameters of Table 4.1 are used. Solid curves: Ψ(1) = 0 dB;

dashed curves: Ψ(1) = -7 dB. SQCP in the desired RL, no SQCP in the interfering

link (’�’). SQCP in both the desired RL and the interfering link (’o’). Performance

upper bound (’�’): SQCP in the desired RL, no interference at all.

73

Practical Interference Mitigation for the Relay Backhaul Link

It is observed from figure 4.2 that SQCP provides notable performance gain

on RL. Here, the curves with ’o’ are plotted using the formulae (4.35) and

(4.36). The interference mitigation gain from SQCP depends heavily on the

strength of the dominant interferer eNB with respect to the sum of interfer-

ence from other eNBs. That is, if the interference experienced by RN on RL

is only from the dominant interferer, then gains will be significant especially

in the low SINR region. For instance, when comparing solid curves with ’o’

and ’�’, the gain is 6 dB at CDF level 0.1 for the dominant interferer only case.

This is clearly higher gain than in the case, where the dominant interferer

represents only half of the total interference, i.e., gain drops to 2-3 dB. It is

noted that in the case with Rice interference but without SQCP, curves were

obtained through simulations. The analytical formulae were not deduced for

this scenario because it was used only for the benchmarking purposes.

Scenario 2: Rayleigh fading - Rayleigh fading

Figure 4.3 presents the CDF of RL SINR when utilizing the FQCP and param-

eters of Table 4.1. It is noted that as an upper bound the case is used where

FQCP is applied in the desired RL and there is no interference at all as denoted

by a dotted curve with ’�’, obtained by using equations (4.37) and (4.38).

-5 0 5 10 15 200

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Figure 4.3. CDF of the SINR of RL when Rayleigh fading occurs in both the desired RL and the

interfering link, and parameters of Table 4.1 are used. Solid curves: Ψ(1) = 0 dB;

dashed curves: Ψ(1) = -7 dB. FQCP in the desired RL, no FQCP in the interfering

link (’�’). FQCP in both the desired RL and the interfering link (’o’). Performance

upper bound (dotted curve with ’�’): FQCP in the desired RL, no interference at

all.

Curves with ’o’ show the performance when FQCP is used in both the desired

74

Practical Interference Mitigation for the Relay Backhaul Link

RL and the interfering link. It is noted that equations (4.39) and (4.40) were

used to generate the curves. Similar to the Scenario 1, it is observed that if

interference experienced on RL is from dominant interferer only, then FQCP

interference mitigation gain is large especially in low CDF range. For exam-

ple, SINR gain is 4 dB or even more for CDF level below 0.3. In addition, if

RN experience same amount of interference from both the dominant inter-

ferer and other interfering eNBs (i.e., Ψ(1) = -7 dB), then the gain from FQCP

interference mitigation is much smaller.

Scenario 3: Rice fading - Rayleigh fading

In this scenario, RN experiences Rice fading in the desired RL and Rayleigh

fading in the interfering links. As before, we apply the upper bound (denoted

by ’�’), where SQCP is employed in the desired RL and there is no interference

at all. It is observed from Figure 4.4, that FQCP provides notable gains while

mitigating the interference on RL.

While the efficiency of SQCP mainly depends on the strength of the static part

in the interference as shown previously in the Figure 4.2, FQCP tracks the fast

fading but needs more feedback. The curves in Figure 4.4 were generated by

using formulae (4.35) and (4.36), where in the parameter c has been computed

from equation (4.41).

0 2 4 6 8 10 12 14 16 18 200

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Figure 4.4. CDF of the SINR of RL when Rice fading occurs in the desired link and Rayleigh

fading in the interfering link, and parameters of Table 4.1 are used. Solid curves:

Ψ(1) = 0 dB; dashed curves: Ψ(1) = -7 dB. SQCP in the desired RL, no feedback in

the interfering link (’�’). SQCP in the desired RL and the FQCP in the interfering

link (’o’). Performance upper bound (’�’): SQCP in the desired RL, no interference

at all.

75

Practical Interference Mitigation for the Relay Backhaul Link

Scenario 4: Rayleigh fading - Rice fading

It is observed from Figure 4.5 that SQCP effectively mitigates the interference

if there is only dominant interferer on the RL: The gap towards the perfor-

mance upper bound (dotted curve with �) becomes small. We used equa-

tions (4.42)-(4.43) to obtain curves.

-5 0 5 10 15 200

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Figure 4.5. CDF of the SINR of RL when Rayleigh fading occurs in the desired RL and Rice

fading in the interfering link, and parameters of Table 4.1 are used. Solid curves:

Ψ(1) = 0 dB; dashed curves: Ψ(1) = -7 dB. FQCP in the desired RL, no feedback in

the interfering link (’�’). FQCP in the desired RL and the SQCP in the interfering

link (’o’). Performance upper bound (’�’): FQCP in the desired RL, no interference

at all.

4.5.2 End-to-end Outage Rate

This section presents how the RL gains impact on the e2e data rate perfor-

mance. We have obtained these curves by using formula (4.44) and SINR dis-

tributions from Section 4.4. We assume Rayleigh fading in the AL with mean

power γa = 8 dB in addition to parameters of Table 4.1. Furthermore, we al-

locate to the AL and RL bandwidth with 180kHz granularity as in LTE where

frequency-time PRBs of (180 kHz)x(1 msec) are used.

76

Practical Interference Mitigation for the Relay Backhaul Link

0 0.2 0.4 0.6 0.8 1 1.20

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Figure 4.6. The e2e outage probability when Rayleigh fading occurs in the AL (γa = 8 dB),

Rice fading in the RL (γr = 12 dB) with Rice interference (γi = 6 dB, Ψ(1) =0 dB).

Numbers of PRB’s are Nr,PRB = 2 and Na,PRB =2 (’o’), 8 (’�’) . Solid blue curves

refer to the case where SQCP is applied in both the desired RL and the interfering

link, dashed red curves refer to case where SQCP is applied only in desired RL.

Figure 4.6 presents the e2e outage rate when Rice fading occurs on RL and two

PRBs are allocated to the RLNr,PRB = 2, while the number of AL PRBsNa,PRB

varies from 2 PRBs to 8 PRBs. It is noted that the number of PRB’s allocated

to RL is usually limited because DeNB may serve many RNs as well as direct

connections simultaneously. In contrast, RN can utilize more radio resources

on AL because RN serves smaller area than DeNB but operate on the whole

frequency band. Hence, in order to boost the RL performance, we employ the

SQCP in the desired RL and the interfering link.

It is noted in the case Na,PRB = 2, AL is the bottleneck and applying the in-

terference mitigation in RL, will improve e2e rate only if outage probability is

high. Yet, if the AL resources increased (i.e., Na,PRB = 8), then the bottleneck

is RL and mitigating the interference on RL will clearly improve the e2e data

rate. At 10% outage probability level, it is observed that rate increases from

0.3 Mbps to 0.5 Mbps in this example.

Figure 4.7 presents the e2e outage rate for scenario where Rayleigh fading

occurs on both the RL and AL (with Rayleigh fading interference in the RL).

As in case of Rice fading AL is bottleneck when Na,PRB = 2 and the gains

from interference mitigation are small. If Na,PRB = 8 for AL, then mitigating

the RL interference will clearly improve the e2e performance. At 10% outage

probability level rate increases from 0.2 Mbps to just over 0.3 Mbps.

77

Practical Interference Mitigation for the Relay Backhaul Link

0 0.2 0.4 0.6 0.8 1 1.20

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Figure 4.7. The e2e outage probability when Rayleigh fading occurs in the AL (γa = 8 dB),

and in the RL (γr = 12 dB) with interference (γi = 6 dB, Ψ(1) =0 dB). Numbers of

PRB’s are Nr,PRB = 2 and Na,PRB =2 (’o’), 8 (’�’) . Solid blue curves refer to the

case where FQCP is applied in both the desired RL and the interfering link, dashed

red curves refer to the case where FQCP is applied only in the desired RL.

4.6 Conclusion

Interference in the RL may severely deteriorate the e2e performance of a prac-

tical dual-hop half duplex DF relaying system. This type of relaying is already

standardized in 4G LTE and will be evidently part of 5G as well. In mobile

systems RNs are preferably employed in the macro-cell edge to improve the

network coverage. Hence, RL can easily become a bottleneck due to interfer-

ence from adjacent macro eNBs. In order to mitigate the impact of interfer-

ence we proposed a simple and practical approach where the signal from the

dominant interfering eNB is mitigated by applying a few bit channel feedback

from the RN to the interfering eNB. The proposed feedback method is already

standardized in 4G to enhance the signal strength in the direct link between

the eNB and UE.

To that end, we derived analytical expressions for the CDF of SINR and the

outage probability on RL by assuming the Rice and Rayleigh fading combina-

tions for the dedicated link and the interfering link. The obtained analytical

results are based on very recent results for Marcum Q-function that enable our

analysis of precoded signal over interfered Rice fading channel. In addition,

we deduced outage probabilities for different interference and feedback sce-

narios using the obtained SINR distributions. Numerical simulations were

78

Practical Interference Mitigation for the Relay Backhaul Link

utilized to verify the analytical results. It was observed that the mitigation

of the dominant interferer on RL enables significant performance gains espe-

cially if the RL creates a bottleneck for the e2e performance. Moreover, for the

Rice fading case with notable K-factor, the required feedback capacity is very

small while gains can be large. The said method is suitable for the infrastruc-

ture relaying where RN can be placed over a rooftop or on a location with LoS

connectivity towards the macro eNBs.

79

Practical Interference Mitigation for the Relay Backhaul Link

80

5. Rapidly Deployable Relays for IndoorEnvironments

5.1 Background and Motivation

5.1.1 Background and motivation

Early mobile networks were designed for voice services and networks were

- in principle - dimensioned to enable an acceptable indoor voice cover-

age [235–237]. Yet, the importance of indoor coverage and capacity were sig-

nificantly underestimated. Nowadays 85% to 90% of mobile services are spent

indoors [238,239], the role of indoor data services has become crucial [239,240]

and practically everybody use mobile devices (e.g., global mobile subscription

reached 7.5 billion in Q3 2016 [241]).

The indoor cellular coverage degrade due to penetration loss experienced by

radio signals obstructing through concrete building floors, walls and mod-

ern window glasses [237, 242–245]. To that end, a number of solutions are

being explored to address indoor coverage problems. Solutions include high

transmission power macro cells that adjust different parameters (e.g., antenna

elevation and downtilt, carrier frequency, etc.) [245–247]. Similarly, pico cell

deployments [248, 249], distributed antenna system (DAS) [250–252], leaky-

feeder installation in tunnels [253, 254] and small cells/femtocells deploy-

ments [255,256] have been introduced to improve indoor coverage. Some ref-

erences are summarized in Table 5.1.

Table 5.1. Legacy solutions for enhancing the indoor wireless coverage

Technology Macro Cell Pico Cell FemtocellDAS & Leaky

Feeders

References [245–247] [248, 249] [255, 256] [250–254]

While technology for indoor coverage provision has been developing fast,

81

Rapidly Deployable Relays for Indoor Environments

there are some serious challenges especially from public safety perspective.

In many indoor locations mobile system capacity is still low making it diffi-

cult e.g. for emergency responders to use video services indoors. Even voice

calls might be blocked due to increased local network traffic. Also, in case of

various disasters local network (base stations and backbone lines) might be

damaged making services unavailable [257–259]. Yet, for emergency respon-

ders it is critically important [260], to quickly establish the communication

link since it significantly enhances the effectiveness of rescue operations and

enables real time critical information sharing [261–263]. Unfortunately, exist-

ing indoor solutions (mentioned in Table 5.1) may be either unavailable (e.g.

due to damage in the emergency) or incapable of providing the QoS and ro-

bustness required emergency operations.

A rapidly deployable relaying network provides an attractive solution for the

improved coverage in several public safety scenarios [158]. Relaying enables

wireless backhaul, prompt and flexible deployment on demand [153] (i.e., no-

madic/moving RN in dynamic radio access networks [152]) and temporary

coverage solution that can meet the public safety requirements. Moreover, a

rapidly deployable RN with exclusive frequency resources may become use-

ful when there is network congestion due to repeated call attempts by the

affected/concerned individuals.

The remaining part of the chapter is organized as follows. Section 5.2 will

provide a brief background literature review of previous contributions in the

context of rapid deployments. Sections 5.3 and 5.4 explain the system models

and simulation parameters and Section 5.5 presents the performance evalua-

tion along with description of simulation results for each deployment option.

Finally, Section 5.6 concludes the study.

5.2 Previous work and Contributions

In literature, several rapidly deployable network prototypes are being stud-

ied as summarized in Table 5.2. For example, an adhoc network is proposed

in [159] deploying portable RN devices which act as a backbone to extend the

wireless coverage in a given incident area. In [160], an airborne wireless sen-

sor networks with emergency communication network is proposed, capable

of sending the GSM warning text messages to the people located in the dis-

aster areas. Wan-Yi et al. proposed in [264] a WiMAX pico base station based

network which can provide cellular coverage to the emergency responders on

urgent cases. In [255, 256], authors proposed solutions for providing a rapid

82

Rapidly Deployable Relays for Indoor Environments

indoor broadband services by exploiting the existing user-deployed closed ac-

cess residential small cells. The work done in [265] propose an airborne 4G

enabled light weight base station that is sent to high altitudes to provide a 4G

coverage in the emergency area.

In [266], a heterogeneous rapid deployment has been presented: Rapid Emer-

gency Deployment mobile Communication (REDComm). In REDComm sev-

eral communication access technologies are aggregated on same platform

providing communication services to all type of users in disaster areas. Sev-

eral REDComm nodes are connected with each other via a mesh network of

802.11a cognitive radio technology operating in unlicensed TV white space

spectrum. The REDComm deployment is dependent on the satellite for con-

nectivity with the outside world. The investigation done in [155] employs the

static and mobile RN to provide link between disconnected nodes and the

nodes located in coverage area, in order to provide cellular coverage in a dis-

aster area. Mentioned work mainly aims to provide alternate communication

links between the network nodes in order to enhance the network sustainabil-

ity in disaster scenarios.

Authors in [267] developed prototype network: Movable and Deployable Re-

source Unit (MDRU), operating in natural disaster area to provide cellular cov-

erage. This system relies on satellite communication to connect with the core

network. EmergeNet, a rapidly deployable cellular type network proposed

in [268] which enables free voice calling and messaging services to the first

emergency responders in the disaster situations. Moreover, the work in [269]

proposes that a multi-hop wireless networks is extended by mobile devices.

This system is known as on-the- fly establishment of multi-hop wireless access

networks (OEMAN). In OEMAN, several mobile devices extend the cellular

coverage to the people present in the emergency locations.

Table 5.2. Proposed and existing rapidly deployable network solutions

Reference Prototype Network Type

[160] WSN Emergency Communication Network (ECN)

[264] WiMAX Picocell Pico Network

[255,256] Femto Cell Femto Cell Network

[265] Helikites Airborne 4G enabled Light Weight Base Station

[266] REDComm Rapid Heterogeneous type Deployment

[267] MDRU Standalone Network

[268] EmergeNet Cellular type Network

[269] OEMAN Extended Cellular type Network

[155,159] Relay Device Adhoc Network

83

Rapidly Deployable Relays for Indoor Environments

Below are some of the benefits of relaying in this context:

Easy deployment/Low cost

Unlike the wireline nodes (e.g., macro, pico, femtocells, etc) as mentioned in

Table 5.1, RN can be connected to the network in almost all locations to pro-

vide local coverage and improved throughput. RN can be designed to be self-

configurable and if highly directive antennas or high transmission powers are

not needed, then cost of the design can be kept low.

Wireless relay backhauling

Unlike the aforementioned prototypes (summarized in Table 5.2) [266, 267],

RN can exploit different RL options to connect the users with the core net-

work. That is, RL can be carried out via macro base stations, using a satellite

link or even via an airborne relay.

Mobile/Nomadic character

The mobile and nomadic nature of RN [152] can be exploited to enable a tem-

porary coverage. The future relaying technologies are expected to be highly

dynamic, energy efficient [151,156] and flexible. Nomadic RN can be mounted

on a vehicle and deployed in the close proximity of a building. This will mini-

mize e.g. the penetration losses when the signal transmission is crossing from

outdoor-to-indoor [17, 245].

This chapter considers rapidly deployable RN used to enable reliable connec-

tivity for the end users in the indoor environments. Therefore, the special

focus is in outdoor-to-indoor relaying. As performance indicators we use the

system throughput and outage probability. We note that several studies has

been done on the feasibility of relaying system for improving the indoor cover-

age. For example, in [17], the network densification was used with infrastruc-

ture RN to enhanced the indoor system capacity. Similarly, relaying indoor

performance evaluation was also done in [13, 14]. Some of the outdoor-to-

indoor relaying results reported in this chapter are also published in [14].

DeNB RN

Figure 5.1. Relaying outdoor-to-indoor coverage for an UE located inside the building

84

Rapidly Deployable Relays for Indoor Environments

5.3 Description of the relay deployment cases

The study of outdoor-to-indoor relaying is carried out for three different case

studies or building designs, namely as 5×5 grid synthetic building used in

3GPP indoor simulation studies [270], dual strip synthetic building model

also used in 3GPP studies [270], realistic building design captured within the

WinProp radio propagation modelling environment [271].

These three case studies described further in the remainder of this sub-section.

5.3.1 5×5 Grid deployment case

The 5×5 apartments layout is one of the 3GPP dense-urban models [270]. This

model consists of 25 square apartments with each size of 10×10 m as shown

in Figure 5.2. We consider only one floor of the building. Building is located

either in the macrocell center or on the edge of the macrocell. The overlaying

macro network consists of seven sites controlled by eNBs. Each site employs

three antennas to provide coverage to three sectors. Each sector contains 10

randomly located outdoor UEs. Eight indoor UEs are deployed inside the

building so that there is at most one UE inside each apartment. We assume

one RN located outside the building 30 m - 60 m away from the external wall

of the building providing the outdoor-to-indoor coverage.

eNB

Relay Node

Outdoor UE

Indoor UE

Tri-Sectored Hexagonal Cellular Network 5 x 5 Grid Building at Cell Edge

0 5 10 15 20 25 30 35 40 45 500

5

10

15

20

25

30

35

40

45

50

60m, 30m

Building at cell center Building at cell edge

Figure 5.2. Relay enhanced cellular network with 5×5 apartment model

5.3.2 Dual Strip deployment case

The 3GPP dual strip model is composed by two blocks of apartments as shown

in Figure 5.3 [270]. Each strip includes two rows of apartments and strips

are separated by a street of 10 m width and they are three floors high. We

85

Rapidly Deployable Relays for Indoor Environments

assume only one dual-strip block located either on the macro cell center or

on the macro cell edge. Network consists of seven eNB sites. Eight indoor

UEs are randomly deployed in strips such that there is at most one UE in

each apartment. We assume two different locations for the relay node. These

locations are: 25m away from the external wall and at the middle of two strips,

see Figure 5.3.

eNB

Relay NodeOutdoor UEIndoor UE

Tri-Sectored Hexagonal Cellular Network

Building at cell center Building at cell edge

Dual Strip Building Model(Three-Storey)

RN located beside

Indoor UEs

25m

10m

RN located at middle

10m

10m

Figure 5.3. Relay enhanced cellular network with 3GPP dual strip building model

5.3.3 Realistic deployment case

In this case we employ the building model that is available within the Win-

prop Software suite [271], see Figure 5.4. The Winprop suite is a software tool,

which includes ray tracing capabilities for modeling of multipath propagation

in different indoor and outdoor realistic environments. Ray tracing technique

takes into account all scattering and reflecting surfaces of each of the signal

rays in the environment.

For this study, the ray tracing method utilized within WinProp is the domi-

nant path model, which provides required trade-off between computational

complexity and accuracy of path loss predictions. The WinProp tool also en-

ables the penetration losses of different building materials (walls, glass, doors,

floors etc.) to be specified explicitly from a provided materials database. For

this case study, we consider an urban area with multiple buildings and a mo-

bile network consisting of four tri-sectored macro sites. From this area we

focus on one particular building as location targeted for outdoor-to-indoor

relaying to provide connectivity to 8 UEs located within this building (see

Figure 5.4). Three candidate relay locations considered (two outdoor and

86

Rapidly Deployable Relays for Indoor Environments

one indoor for comparison) as shown in Figure 5.5.

Location 1

Location 2

Indoor LocationeNB 1

eNB 2

eNB 3eNB 4

Tri-Sectored 4 eNBs relaying enhanced cellular network

Tri-sectored eNB Relay Node

Figure 5.4. Relay enhanced cellular network with realistic building model.

Location 1

Location 2

Indoor Relay

eNB 1

Figure 5.5. Relay locations around the proposed building.

5.4 System Model

We investigate the system performance for ideal and non-ideal RL. The ideal

RL does not restrict the AL capacity and relates to the case where RN has

particularly good channel conditions and frequency resources are not setting

capacity limitations. That is, RN with ideal RL is comparable to a pico node

deployment. While in case of non-ideal RL, the RL capacities could be bottle-

neck for the AL capacities.

87

Rapidly Deployable Relays for Indoor Environments

Table 5.3 summarizes the RL scenario, scheduling scheme and propagation

model combinations in the forthcoming study. Scheduling, channel models

and other system parameters are described in more details in the following

sections.

Table 5.3. Ideal and non-ideal relay backhauling

RL Backhauling Deployment Scheduling Schemes Propagation Models

Ideal RL 5×5 Grid Round Robin (RR) 3GPP

Non-Ideal RL

5×5 Grid

Dual Strip

Realistic

Max-Min Fairness (MMF) COST-231 WI

5.4.1 Throughput Model and Radio Resource Scheduling

In case of non-ideal RL we employ the resource scheduling strategy as shown

in Figure 2.13. There the DeNB allocates three (MBSFN) subframes out of the

ten to RL while the remaining seven subframes are shared by DeNB and RN

to serve UEs. The amount of data transferred over the RL (per 10 msec radio

frame) is denoted by Dr:

Dr = NMBSFN ·NPRB ·BPRB · Sr, (5.1)

where NMBSFN refers to the number of MBSFN subframes, NPRB denotes the

number of physical recourse blocks per subframe andBPRB denotes the band-

width of a PRB. The Sr is approximated from Υr via the modified Shannon’s

formula which is adjusted by two parameters namely bandwidth efficiency (

Beff ) and SINR efficiency (Υeff ):

Sr = Beff · log2(1 +

Υr

Υeff

). (5.2)

Here, Υr =PRx,DeNB

Pn +∑

PRx,Other eNBs, PRx,DeNB is the desired received power by

RN from DeNB and PRx,Other eNBs is the power received by the RN from other

neighbouring eNBs. Similarly, the transferred data over the AL is given by

Da =U∑

u=1

Da,u =U∑

u=1

NASF ·NPRB,u ·BPRB · Sa,u, (5.3)

where U refers to the number of UEs served by RN, Da,u is the individual

throughput of a uth user, NASF is the number of subframes allocated to AL

in a 10 msec frame, NPRB,u denotes the number of physical recourse blocks

allocated to uth user and Sa,u is the spectral efficiency on AL of uth user. The

spectral efficiency for an UE with Υa,u is given by

Sa,u = Beff · log2(1 +

Υa,u

Υeff

). (5.4)

88

Rapidly Deployable Relays for Indoor Environments

We employ the Max-Min fairness (MMF) scheduling in the AL. This scheme

aims to enhance the throughput performance of indoor UEs, experiencing

the worst SINR levels. The worst indoor UEs are being allocated with more

PRBs so that all the UEs achieve the same throughput. MMF is especially

suitable in public safety scenarios if there is no prioritization between different

(authorized) users. If some rescue teams or actors in the field are prioritized

over others, then MMF needs to be modified accordingly.

The number Na of allocated resource blocks in AL is limited such that Na =

NASF · ∑Uu=1NPRB,u ≤ Nmax. We also assume that AL throughput Da per

each 10 msec frame is not exceeding the RL capacity Dr. That is, there is no

buffer in the RN. We require that for the individual UE throughput Da,u there

holds Da,u ≥ Dmin. The scheduling is described in Algorithm 1. We assume

an initial maximum and minimum data rate values, i.e., D0 and Dmin, respec-

tively, for the AL and direct link. Then D0 is decreased with 100 kbps per step

if the resource allocation fails with current D0.

Algorithm 1 Max-Min fairness scheduling

1: Compute the RL throughput Dr per 10 msec using (5.1)

2: Assume the initial value of D0

3: Compute the individual throughputs for UEs such that Da,u ≥ D0,

4: if Na ≥ Nmax then

5: Decrease the value of D0 and proceed to 3.

6: if D0 < Dmin then

7: Drop the worst indoor UE.

8: end if

9: end if

5.4.2 Propagation Models

We employ the 3GPP and COST 231 Walfisch-Ikegami channel models for

the path loss estimation in the direct link, RL and AL, see Table 5.4. There,

S denotes the distance in kilometers between the transmitter and receiver

and fc represents the carrier frequency in MHz. We note that in 3GPP mod-

els of Table 5.4 path loss is computed from formula Prob(S)PLLoS + (1 −Prob(S))PLNLoS.

89

Rapidly Deployable Relays for Indoor Environments

Table 5.4. COST-231-WI and 3GPP channel models assumed in simulations

Links Propagation Models

Direct Link COST231−WI PLNLoS = 42.497 + 38 · log10(S) + 24.5 · log10(fc) + 0.00162 · fc · log10(fc)Relay Backhaul Link COST231−WI PLNLoS = 34.539 + 38 · log10(S) + 24.5 · log10(fc) + 0.00162 · fc · log10(fc)Access Link COST231−WI PLNLoS = 42.6 + 26 · log10(S) + 20 · log10(fc)Direct Link 3GPP Prob(S)Urban = min(0.018S , 1) · (1− exp(

−S0.063

)) + exp(−S

0.063)

Prob(S)Suburban = exp−(S−0.010.2

)

PLLoS = 103.4 + 24.2 · log10(S)PLNLoS = 131.1 + 42.8 · log10(S)

Relay Backhaul Link 3GPP Prob(S)Urban = min(0.018S , 1) · (1− exp(−S

0.072)) + exp(

−S0.072

)

Prob(S)Suburban = exp−(S−0.010.23

)

PLLoS = 100.7 + 23.5 · log10(S)PLNLoS = 125.2 + 36.3 · log10(−S)

Access Link 3GPP Prob(S)Urban = 0.5−min(0.5, 5 · exp(−0.156S

)) + min(0.5, 5 · exp( −S0.03

))

Prob(S)Suburban = 0.5−min(0.5, 3 · exp(−0.3S

)) + min(0.5, 3 · exp( −S0.095

))

PLLoS = 103.8 + 20.9 · log10(S)PLNLoS = 145.4 + 37.5 · log10(S)

5.4.3 Antenna pattern

The antenna gain depends on the antenna pattern of a transmitter. Antenna

pattern is a 3D graphical representation of antenna radiation properties as

a function of direction. The 3GPP pattern for a directional antenna is given

by (5.5) [173, 272]:

A(θ) = −min

[12

θ3dB

)2

,Am

], (5.5)

where A(θ) denotes the antenna gain at angle θ such that −180o ≤ θ ≤ 180o

and θ3dB denotes the angle of direction where gain is 3 dB lower than in the

main direction. The antenna front-to-back ratio is denoted by Am and it has

value 25 dB and 20 dB for macro eNB and relay node antennas, respectively.

5.4.4 System parameters and assumptions

We focus on the downlink of 3GPP Case 1 (Urban) and Case 3 (Suburban)

macro-cell layouts with inter-site distances (ISD) 500 m and 1732 m respec-

tively. For the current study, we ignore the interference received on RL from

the neighboring eNBs. Table 5.5 presents the simulation parameters for all

three deployment cases.

90

Rapidly Deployable Relays for Indoor Environments

Table 5.5. Simulation Parameters

Parameters 5×5 Grid Dual Strip Realistic

Ideal RL Non-Ideal RL Non-Ideal RL Non-Ideal RL

Air Interface LTE-FDD LTE-FDD LTE-FDD LTE-FDD

Carrier Frequency 2000 MHz 800 MHz 2000 MHz 400 MHz

Operational BW 10 MHz 10 MHz 10 MHz 5 MHz

Fading Log-Normal Shadowing, WinProp [271], Rayleigh/Rician

Propagation 3GPP COST-231 WI 3GPP Winprop [271]

Scheduling RR MMF MMF MMF

Standard deviation 8 dB for Direct Link, 6 dB RL, 10 dB AL –

Penetration loss 20 dB 0.6 dB/m 0.6 dB/m –

Thermal Noise PSD -174 dBm/Hz -174 dBm/Hz -174 dBm/Hz -174 dBm/Hz

BW Efficiency 0.88 0.88 0.88 0.88

SINR Efficiency 1.25 1.25 1.25 1.25

IL 5 dB 5 dB 5 dB 5 dB

Macro Parameters

Tx Power 46 dBm 46 dBm 46 dBm 46 dBm

Ant Pattern A(θ) = - min

[12

θ3dB

)2

,Am

]θ3dB = 70 ◦, Am = 25 dB Directional

Ant Elevation 25 m 25 m 25 m Table 5.6

Ant Configuration Tx-2, Rx-2 Tx-2, Rx-2 Tx-2, Rx-2 Winprop [271]

eNB Ant Gain 14 dBi 14 dBi 14 dBi –

Intersite Distance 500 m / 1732 m 500 m 500 m –

Diversity Gain 5 dB 6 dB 6 dB 6 dB

Relay Node Parameters

Tx Power 30 dBm 30 dBm 30 dBm 30 dBm

Ant Pattern Omni & Directional Omni Omni

Ant Elevation 5 m 5 m 5 m Table 5.6

Ant Configuration Tx-1, Rx-2 Tx-1, Rx-2 Tx-1, Rx-2 Winprop [271]

Diversity Gain 5 dB 6 dB 6 dB 6 dB

UE Parameters

Height/Location UE height varies from 1.5 m to 7 m

Noise Figure 9 dB 9 dB 9 dB 9 dB

UE Number 10 Outdoor and 8 Indoor LTE-UEs

Ant Configuration Tx-1, Rx-2 Tx-1, Rx-2 Tx-1, Rx-2 Tx-1, Rx-2

91

Rapidly Deployable Relays for Indoor Environments

Table 5.6. Antenna heights in realistic deployment case

eNB/RN Height (m)

eNB 1 Ant 1 50

eNB 1 Ant 2 40

eNB 1 Ant 3 40

eNB 2 Ant 1 25

eNB 2 Ant 2 25

eNB 2 Ant 3 25

eNB 3 Ant 1 25

eNB 3 Ant 2 25

eNB 3 Ant 3 25

eNB 4 Ant 1 50

eNB 4 Ant 2 50

eNB 4 Ant 3 50

Indoor RN 1

Outdoor RN 10

5.5 Performance Evaluation and Simulations

This section presents the simulation results generated in the selected deploy-

ment cases. Network performance is explained in terms of SINR per PRB and

indoor UEs throughput.

5.5.1 5× 5 Grid deployment case

We consider two cases: RL is ideal and RL is non-ideal. Furthermore, we carry

out a comparative performance evaluation when RN is deployed at 60 m and

30 m away from the center of the building.

Ideal RL

Here we assume the RR scheduling in AL and direct link. Figure 5.6 shows

the outdoor-to-indoor relaying performance for different RN locations. The

eNB only network case has been presented for the sake of benchmarking. Fig-

ure 5.6 is comprised of four sub plots which show the CDFs of the SINR per

PRB (a & c) and indoor UE throughput (b & d). The performance analysis has

been done for scenarios where building is on the macro-cell center and on the

macro-cell edge.

92

Rapidly Deployable Relays for Indoor Environments

Figure 5.6. REC performance when RN is located 60 m and 30 m away from the center of the

building, (a) SINR per PRB (Cell Center), (b) Indoor UE throughput (Cell Center),

(c) SINR per PRB (Cell Edge), (d) Indoor UE throughput (Cell Edge),

As expected, RN deployment deteriorate the SINR levels of indoor UEs espe-

cially when building is at cell center due to high interference received from

nearby high power transmission of eNBs as shown in Figure 5.6 (a). However,

SINR is considerably improved when the RN-Building distance is reduced

(i.e., when it is 30 m away from the building center) as compared to the 60 m

separation. Table 5.7 presents SINR improvements for different RN locations

in both ISD=500 m and ISD=1732 m cases.

Although SINR is decreased due to RN deployment, it considerably improves

the indoor UE throughput as compared to the eNB only case. This follows

from the fact that resources in AL are less competed than in the direct link

where eNB also serves outdoor UEs. The indoor UE throughput are shown

in Figure 5.6 (b) and (d). If building is on the cell edge, then RN provides good

performance gain as compared to the eNB only case. From Figure 5.6 (c) we

find that SINR is degraded less in case of macro-cell edge deployment than

it is in the macro-cell center case. These SINR gains along with reduced the

contention for RN radio resources translate to the enhanced UE throughput.

93

Rapidly Deployable Relays for Indoor Environments

Table 5.8 shows the throughput gains of relaying.

Table 5.7. Achievable SINR for indoor relay UEs for different RN locations and ISDs

SINR Enhancements (dB)

ISD 500 m

RN-Building DistanceCell Center Cell Edge

10th %-ile 50th %-ile 80th %-ile 10th %-ile 50th %-ile 80th %-ile

30 m 0.5367 5.6279 11.1129 0.3053 6.1023 11.8682

50 m -0.4250 3.9264 8.4634 -0.7235 3.9931 8.4943

60 m -0.5459 3.6054 8.2098 -1.1602 3.1696 7.4261

70 m -0.7474 3.1359 7.9763 -1.5920 2.4297 6.5254

ISD 1732 m

10th %-ile 50th %-ile 80th %-ile 10th %-ile 50th %-ile 80th %-ile

30 m 10.8367 22.5908 29.8260 18.0678 27.9942 33.5150

50 m 8.3682 18.8225 25.4026 14.4898 23.9955 29.2661

60 m 7.3320 17.0700 23.8905 12.7769 22.5130 27.5884

70 m 6.4327 15.7102 22.5332 10.9306 21.0087 26.1429

Table 5.8. Achievable throughput for indoor relay UEs for different RN locations and ISDs

Throughput Enhancements (Mbps)

ISD 500 m

RN-Building DistanceCell Center Cell Edge

10th %-ile 50th %-ile 80th %-ile 10th %-ile 50th %-ile 80th %-ile

30 m 0.5696 1.6394 3.3166 0.5442 1.6243 3.2905

50 m 0.4299 1.3861 2.7970 0.3854 1.3124 2.7889

60 m 0.3972 1.2072 2.5243 0.3457 1.1712 2.6416

70 m 0.3847 1.1206 2.1578 0.3148 1.0156 2.3741

Figure 5.7 presents results when RN utilizes either omni-directional or direc-

tional antennas for the transmission on AL. It is noted that the use of direc-

tional antenna provides notable performance gain in terms of SINR (a) and

throughput (b) as compared to the omni-directional antenna case as evident

in Tables 5.9 and 5.10. This is because the directional antenna concentrate the

relay transmission to the indoor RUEs as well as decrease the interference to-

wards the indoor UEs served by eNB as shown in Figure 5.7 (a). The improved

SINR levels are directly translate to enhanced throughput as can be seen from

Figure 5.7 (b). Tables 5.9 and 5.10 present the performance enhancements due

to directional antenna on AL.

94

Rapidly Deployable Relays for Indoor Environments

Figure 5.7. SINR and throughput when building is located at the macro-cell edge and RN is 60

m away from the center of the building. Omni-directional and Directional antennas

are used in the AL transmission, (a) SINR per PRB and (b) Indoor UE, MUE and

RUE throughput.

Table 5.9. SINR gains by using directional Antennas for AL transmissions with ISD 500 m and

RN distance of 60 m

Impact of omni-directional and directional antennas on AL SINR (dB)

User Equipment

Cell Center

Omni Directional

10th %-ile 50th %-ile 80th %-ile 10th %-ile 50th %-ile 80th %-ile

Indoor UE -0.5459 3.6054 8.2098 -0.0533 4.3623 9.2467

Indoor RUE -0.5459 3.1447 7.2116 0.0757 4.2252 8.6904

Indoor MUE -0.6627 3.8044 8.9427 -0.1484 4.4230 9.5552

User Equipment

Cell Edge

Omni Directional

10th %-ile 50th %-ile 80th %-ile 10th %-ile 50th %-ile 80th %-ile

Indoor UE -1.1602 3.1696 7.4261 -0.4548 4.5592 9.0919

Indoor RUE -0.7220 3.5206 7.5625 0.0809 5.0851 9.5924

Indoor MUE -1.3477 2.9325 7.3938 -0.7852 4.1930 8.7780

95

Rapidly Deployable Relays for Indoor Environments

Table 5.10. Throughput enhancements by using directional antennas for AL transmissions

with ISD 500 m and RN distance of 60 m

Impact of omni-directional and directional antennas on AL throughput [Mbps]

User Equipment (UE)

Cell Center

Omni Directional

10th %-ile 50th %-ile 80th %-ile 10th %-ile 50th %-ile 80th %-ile

Indoor UE 0.3972 1.2072 2.5243 0.4488 1.5788 3.4511

Indoor RUE 1.0560 2.5398 5.0902 1.2040 2.8811 5.4756

Indoor MUE 0.3699 1.0318 2.0468 0.4003 1.2293 2.5882

User Equipment (UE)

Cell Edge

Omni Directional

10th %-ile 50th %-ile 80th %-ile 10th %-ile 50th %-ile 80th %-ile

Indoor UE 0.3457 1.1712 2.6416 0.3853 1.4946 3.4518

Indoor RUE 0.9178 2.1525 3.9245 1.0613 2.5765 4.7295

Indoor MUE 0.3153 0.9016 2.0731 0.3379 1.0737 2.7921

Non-ideal RL

If RL is not ideal, then we employ the resource scheduling strategy of Fig-

ure 2.13 to share the radio resources among the direct link, RL and AL. That

is, DeNB allocates three LTE subframes out of ten per each radio frame for RL

while the remaining seven subframes are allocated for the direct link in DeNB

and AL in RN.

Figure 5.8. Performance when building is at the macro-cell center, (a) Indoor UE throughput,

(b) Indoor RUE throughput, (c) Indoor MUE throughput, (d) Number of indoor

UEs embraced to macro eNB and RN.

Figure 5.8 presents the indoor UE (for both indoor MUEs and RUEs) through-

put performance when ISD=500 m and building is at the macro-cell center.

96

Rapidly Deployable Relays for Indoor Environments

The RN distance from the center of the building is either 60 m or 30 m. Re-

sults show that the use of an RN clearly improves the indoor UE performance

as compared to eNB only scenario. Moreover, the indoor UE performance

gain is further improved when the RN is deployed closer to the building.

Figure 5.9 presents the impact of subframe allocation to RL. Since AL is the

bottleneck now, the UE performance becomes worst when more resources are

allocated to RL. This is interesting since in most of the deployment options

studied in literature RL represents the end-to-end performance bottleneck.

1 2 3 4 5 6 70

0.5

1

1.5

2

2.5

3 106

Figure 5.9. Indoor UE performance for different RL subframe (MBSFN) allocations.

5.5.2 Dual Strip deployment case

Here we assume two alternative locations for the RN deployment namely, be-

side the building strips and in the middle of two strips as shown in Figure 5.3.

Figure 5.10 presents the indoor UE throughput when building is located at

the macro-cell center. It is observed that RN provides throughput gains in

both deployment scenarios as compared to the eNB only case. It is also noted

that there is more indoor UEs connected to RN, if it is deployed in the middle

of strips as compared to the scenario where RN is deployed beside the build-

ing strips. Moreover, indoor UE throughput are further improved when the

building and RN are located at the macro-cell edge.

97

Rapidly Deployable Relays for Indoor Environments

Figure 5.10. UE throughput when RN is located beside and middle of the building strips and

building is located either at the macro-cell center or at the macro-cell edge, (a)

Indoor UE throughput, (b) Indoor UEs connected to eNB and RN

5.5.3 Realistic deployment case

In this deployment we assume two outdoor RN locations (i.e., location 1 and

location 2) and one indoor RN location as shown in Figure 5.5. To that end,

Figure 5.11 presents the throughput of indoor RUEs and indoor MUEs.

We find from results that indoor UE performance is improved when it is

served by RN located inside the building as compared to the case where RN

is in either of the outdoor locations, see Figure 5.11 (a). Yet, this follows from

the fact that there are less UEs connected to RN if it is located indoors, see Fig-

ure 5.11 (b). Then RN coverage is comparatively good on the deployed floor

as compared to the rest of the floors in the building. Moreover, the number of

UEs connected to RN when it is in the location 1 is higher than in location 2,

because the RN transmission on AL is less interfered by macro eNB in loca-

tion 1. It is also observed that RN experiences good RL at location 2, but still,

its indoor UEs receives high interference from macro eNB and throughput on

the AL are reduced.

Figure 5.11. UE throughput and connection nodes when RN is deployed at outdoor locations 1

or 2, or inside the building, (a) Indoor UE throughput, (b) Number of indoor UEs

connected to eNB and RN

98

Rapidly Deployable Relays for Indoor Environments

5.6 Conclusions

We considered few rapid RN deployment scenarios within the 3GPP LTE-A

Type 1 inband relaying framework. The main objective was to examine the

relaying indoor performance in three different RN deployment cases, namely

in 3GPP 5×5 Grid, in 3GPP Dual-strip and in a realistic deployment cases.

We showed via system level simulations that RN can be used to improve emer-

gency service coverage and capacity in the studied outdoor-to-indoor cases.

The RUEs experience high interference from macro eNB, but still these UEs

experience less competition for the radio resource on AL and throughput is

improved accordingly. It was also observed that the end-to-end UE perfor-

mance depends on the propagation conditions on both RL and AL. The LoS

condition can be obtained due to the flexible and rapid-deployment nature of

RN. Simulation results also show that the indoor UE SINR is enhanced the

closer the RN is deployed to the building. Performance is in all cases better

when the building is located at the macro-cell edge. If RN employs directional

antenna for AL transmission, the system performance is improved because

the RN transmission is then focused towards the dedicated indoor UEs. Di-

rective antenna also minimizes the likelihood of RN interference towards the

UEs served by macro eNB.

Finally, we also addressed the radio resource management (RRM) issues in a

network with outdoor-to-indoor relaying. We considered the resource split-

ting between the RN and UE on direct link at DeNB and employed MMF

scheduling scheme on AL. The MMF scheduling scheme aims to prioritize

the worst indoor UEs. According to the simulation results, worst indoor UEs

performance was considerably improved.

99

Rapidly Deployable Relays for Indoor Environments

100

6. Conclusions and future work

Mobile communication technologies have developed rapidly to meet the chal-

lenges of rapidly increasing data service demands. To that end, relaying is one

of the attractive candidate technologies to fulfill the stringent requirements

set by ITU-R and 3GPP for future mobile communication systems. This the-

sis focused on a two-hop decode-and-forward relaying in a mobile commu-

nication network. Thesis contributions include design, analysis and perfor-

mance evaluation of different radio resource usage approaches between the

two hops (i.e., between Relay Link (RL) and Access Link (AL)). Study also

covers interference mitigation methods that can be used to relax the RL limi-

tation that may become a bottleneck for the end-to-end (e2e) capacity. Finally,

thesis considers the outdoor-to-indoor coverage improvements using rapidly

deployable relays.

6.1 Analysis of Optimal Resource Sharing

The theme of the analysis carried out in Chapter 3 was the rate enhancement

through Resource Allocation (RA) between the RL and AL. In addition, the

performance impact from DL/UL decoupling was investigated.

A comparative analysis of conventional RA, fixed RA with and without buffer-

ing, and (resource) optimal RA were carried out. Closed-form expressions

were derived for the mean and outage e2e data rate of the two-hop DF relaying

system. In case of optimal RA the rate analysis led to an integral presentation

for which a tight (closed-form) lower bound was found.

The obtained closed-form expressions enabled a simple performance evalu-

ation where mathematical formulas were also validated through numerical

simulations. Results show that RA is essential for the performance of DF re-

lay systems. The (resource) optimal RA clearly improves the e2e relaying per-

formance but requires exchange of fast scheduling information. On the other

101

Conclusions and future work

hand, buffer can be used to compensate the lack of fast scheduling informa-

tion and fixed RA with large buffer shows very good performance. Of course

buffering increases the delay and is not necessarily feasible for all service

types. The fixed RA without buffer is feasible if service demand is not chang-

ing fast while conventional RA is clearly inferior to other RA schemes and

benefits only from its implementation simplicity. Finally, it was shown that

decoupling the DL/UL transmissions improve the spectral efficiency. This

approach will be especially beneficial when tackling the imbalance between

small (relay controlled) cells and large macro-cells.

Future work on resource optimal relaying will focus on two aspects: the spe-

cific RRM protocols and the impact of multi-antenna methodologies on the

performance of resource optimal relaying. The current analysis is very generic

and covers the case where the time sharing between RL and AL resources is

done. Yet, e.g. in 4G and 5G radio resources are forming a time-frequency

resource pool and user is assigned resources to satisfy the rate request. If

e2e link needs to carry the same amount of resources over RL and AL (as in

resource optimal relaying), then e2e scheduler should take into account the

resource needs of different users when scheduling them. If rates are not high

and number of users is large, then resource scheduling close to optimal can be

executed but in case users occupy large amounts of time-frequency resources,

it will be challenging to obtain optimal e2e scheduling. Future research will

focus on this aspect. We will, for example, consider how multi-antenna meth-

ods and different fast scheduling approaches can be used to obtain optimal or

close-to-optimal resource allocation over the e2e link.

6.2 Interference Mitigation for the Relay Backhaul Link

This section concludes the work in Chapter 4. There special focus was on the

DL of a wireless relay backhaul, where RN, in addition to the desired signal,

also receives strong interference from the neighbouring eNB. To relax the RL

bottleneck we proposed a simple and practical approach where we apply a

few bit channel feedback to improve SINR from DeNB and to mitigate the

interference from the dominant interfering eNB.

To that end, the contribution was two-fold. First, an analytical approach for

calculating the CDF of SINR and the outage probability in the RL was pro-

posed. The Rice and Rayleigh fading combinations for the desired and dom-

inant interfering links were assumed. Outage probabilities were deduced for

different interference and feedback scenarios based on the obtained SINR dis-

102

Conclusions and future work

tributions. Results showed that the mitigation of the dominant interferer in

the RL leads to notable performance gains especially if the RL represents a

bottleneck for the e2e relaying performance. Second, a simulation campaign

was carried out to study the performance of proposed transmit beamforming

schemes as compared to the baseline scenario. It was shown in LTE frame-

work that the SINR per PRB in RL and accordingly the e2e data rate can be

clearly improved with just few bit channel information that is fed back to the

interfering eNB.

The future work contains further improvements in the backhaul performance

by designing cooperative and interference-aware resource scheduling and an-

tenna schemes to enhance the e2e data rate performance. Especially, it is im-

portant to investigate more thoroughly the impact of intercell interference and

design better cooperative scheduling approaches over adjacent cells. There,

joint scheduler over multiple cells should beneficially utilise the channel feed-

back from users and the scheduling status information from different base

stations. In such scenario multi-objective optimisation will be needed.

6.3 Rapidly Deployable Relays for Outdoor-to-indoor Coverage

The performance evaluation in Chapter 5 focused on the rapidly deployable

relays that can be used in mobile systems to provide a temporary outdoor-

to-indoor coverage in e.g. emergency/public safety situations. Evaluation

considered three different building models to provide indoor environment

diversity. These deployments were: 3GPP 5x5 grid, 3GPP dual strip model

and realistic deployment case where a certain building layout with more re-

alistic (ray-tracing) propagation model was applied.

The objective of the simulation campaign was to examine the impact of

rapidly-deployed outdoor RN on the indoor service provision in terms of e2e

throughput experienced by indoor UEs. Simulation results showed that RN

can significantly improve the indoor system performance. It was observed

that the RUEs served by RN were experiencing heavy interference from the

overlaying macro eNB but they competed less for the abundant AL radio re-

sources. This led to enhanced indoor throughput.

The relaying performance depends heavily upon the propagation conditions

on both RL and AL. The Line-of-Sight towards the covered building can be

achieved due to the nomadic nature of the rapidly deployable relays. Re-

sults indicate the significant impact of the location of outdoor deployed RN

on achievable performance of UEs inside the building. Furthermore, the ben-

103

Conclusions and future work

efits of relaying became especially visible when target building was located

on the macro-cell edge where eNB coverage is weak and macro eNB will not

interfere heavily the relay AL. Moreover, it was also shown that the usage of

directional antenna in RN can provide notable additional gains. When using

so-called max-min scheduling, the relaying was clearly improving the perfor-

mance of the worst indoor UEs. This work will not be continued in future.

104

References

[1] E. Dahlman, S. Parkvall, and J. Skold. 4G, LTE-Advanced Pro and The Road to 5G.

3rd ed., Academic Press, July, 2016.

[2] “Ericsson Mobility Report”, June, 2017 [Online].

Available:https://www.ericsson.com/assets/local/mobility-

report/documents/2017/ericsson-mobility-report-june-2017.pdf.

[3] A. Osseiran, J. F. Monserrat, and P. Marsch. 5G Mobile and Wireless Communica-

tions Technology. Cambridge University Press, UK, 2016.

[4] J. M. C. Brito. “Trends in wireless communications towards 5G networks - The

influence of e-health and IoT applications”. In International Multidisciplinary

Conference on Computer and Energy Science (SpliTech), July, 2016.

[5] D. L. Pérez, M. Ding, H. Claussen, and A. H. Jafari. “Towards 1 Gbps/UE in

Cellular Systems: Understanding Ultra-Dense Small Cell Deployments”. IEEE

Communications Surveys & Tutorials, 17:2078–2101, June, 2015.

[6] M. Kamel, W. Hamouda, and A. Youssef. “Ultra-Dense Networks: A Survey”.

IEEE Communications Surveys and Tutorials, 18:2522–2545, May, 2016.

[7] N. Bhushan, J. Li, D. Malladi, R. Gilmore, D. Brenner, and A. Damnjanovic.

“Network densification: the dominant theme for wireless evolution into 5G”.

IEEE Communications Magazine, 52:82–89, February, 2014.

[8] Cisco, “Cisco visual networking index: Global mobile data traffic fore-

cast update, 2016-2021,” White Paper,February, 2017. [Online]. Avail-

able: https://www.cisco.com/c/en/us/solutions/collateral/service-

provider/visual-networking-index-vni/mobile-white-paper-c11-520862.pdf.

[9] 3rd Generation Partnership Project; Technical Specification Group (TSG) Radio

Access Network, “Evolved Universal Terrestrial Radio Access (E-UTRA); Fur-

ther advancements for E-UTRA physical layer aspects (Release 9),” 3GPP, Tech.

Rep. 36.814 v9.0.0, March, 2010.

[10] I. Ullah, A. Dowhuszko, Z. Zheng, D. González G., and J. Hämäläinen. “End-

to-end data rate performance of Decode-and-Forward Relaying with Differ-

ent Resource Allocation Schemes”. Mobile Information Systems (MSI), 2017:11,

September, 2017.

[11] A A. Dowhuszko, I. Ullah, D. González G, and J. Hämäläinen. “Performance

of decoupled uplink and downlink in a macrocellular network with DF relays

”. Prepared for submission, 2017.

105

References

[12] I. Ullah, E. Mutafungwa, J. Hämäläinen, and D. González G. “A simple ap-

proach for the suppression of Rician interference in a relay backhaul using

limited feedback ”. in Physical Communication, June, 2018.

[13] I. Ullah, B. B. Haile, E. Mutafungwa, and Jyri Hämäläinen. “Use of Beamform-

ing and Interference Mitigation Techniques for Relay Backhaul Enhancement

”. In 39th International Conference on Telecommunications and Signal Processing

(TSP), Vienna, Austria, June, 2016. IEEE.

[14] I. Ullah, Z. Zheng, E. Mutafungwa, and Jyri Hämäläinen. “On the Use of

Nomadic Relaying for Emergency Telemedicine Services in Indoor Environ-

ments”. In 3rd MOBIHEALTH International Conference on Wireless Mobile Com-

munication and Healthcare, volume 61, pages 61–68, Paris, France, November

2012. Springer Berlin Heidelberg.

[15] C. Hoymann, D. Astely, M. Stattin, G. Wikstrom, J. F. Cheng, A. Hoglund,

M. Frenne, R. Blasco, J. Huschke, and F. Gunnarsson. “LTE release 14 outlook

”. IEEE Communications Magazine, 54:44–49, June June, 2016.

[16] “International Telecommunication Union - Radio (ITU-R) Recommendation.

M.2083-0, IMT vision framework and overall objectivs of the future develop-

ment of IMT for 2020 and beyond ”, September 2015.

[17] A. Osseiran, F. Boccardi, V. Braun, K. Kusume, and P. Marsch. “Scenarios for 5G

mobile and wireless communications: the vision of the METIS project”. IEEE

Communications Magazine, 52(02), May, 2014.

[18] Ö. Bulakci, S. Redana, B. Raaf, and J. Hämäläinen. “Performance enhancement

in LTE-Advanced relay networks via relay site planning”. In 71st IEEE Vehicular

Technology Conference (VTC), pages 1–5, May, 2010.

[19] A. B. Saleh, Ö. Bulakci, J. Hämäläinen, S. Redana, and B. Raaf. “Analysis of

the impact of site planning on the performance of relay deployments ”. IEEE

Transactions on Vehicular Technology, 61:3139–3150, June, 2012.

[20] Ö. Bulakci. “Backhaul link enhancement and radio resource management for relay

deployments”. PhD thesis, Aalto University School of Electrical Engineering,

June, 2013.

[21] A. B. Saleh, S. Redana, J. Hämäläinen, and B. Raaf. “Resource sharing in relay-

enhanced 4G networks”. In 11th European Wireless Conference 2011 - Sustainable

Wireless Technologies (European Wireless), April, 2011.

[22] M Minelli, M. Ma, M. Coupechoux, and P. Godlewski. “Scheduling Impact on

the Performance of Relay-Enhanced LTE-A Networks”. IEEE Transactions on

Vehicular Technology, 65:2496 – 2508, April, 2015.

[23] Ö. Bulakci, J. Hämäläinen, and E. Schulz. “Performance of coarse relay site

planning in composite fading/shadowing environments”. In IEEE 24th Interna-

tional Symposium on Personal Indoor and Mobile Radio Communications (PIMRC),

September, 2013.

[24] Ö. Bulakci, A. B. Saleh, J. Hämäläinen, and S. Redana. “Performance Analy-

sis of Relay Site Planning Over Composite Fading/Shadowing Channels With

Cochannel Interference”. IEEE Transactions on Vehicular Technology, 62:1692–

1706, December, 2012.

106

References

[25] S. Yi and M. Lei. “Backhaul resource allocation in LTE-Advanced relaying sys-

tems”. In Proceedings in IEEE Wireless Communications and Networking Confer-

ence, April, 2012.

[26] W. Xuan-li, Z. Wan-jun, and W. Wei. “Throughput and fairness-balanced re-

source allocation algorithm in TD-LTE-Advanced relay-enhanced network”.

In International Workshop on High Mobility Wireless Communications (HMWC),

November, 2013.

[27] D. W. Kifle, Ö. Bulakci, A. B. Saleh, S. Redana, and F. Granelli. “Joint backhaul

co-scheduling and relay cell extension in LTE-advanced networks uplink per-

formance evaluation”. In 18th European Wireless Conference (European Wireless),

April, 2012.

[28] A. B. Saleh, Ö. Bulakci, S. Redana, B. Raaf, and J. Hämäläinen. “A divide-and-

conquer approach to mitigate relay-to-relay interference”. In IEEE 22nd Interna-

tional Symposium on Personal Indoor and Mobile Radio Communications (PIMRC),

September, 2011.

[29] M. Eguizábal and Á. Hernández. “Dynamic, fair and coordinated resource al-

location for backhaul links for heterogeneous load conditions in LTE-advanced

relay systems”. In 6th International Congress on Ultra Modern Telecommunications

and Control Systems and Workshops (ICUMT), October, 2014.

[30] B. Park, H. Lee, D. Park, and K. Kwon. “LTE-Advanced relay system perfor-

mance with backhaul link enhancement”. In 4th IEEE International Conference

on Communication and Electronics (ICCE), pages 206–211, August, 2012.

[31] Y. N. R. Li, H. Xiao, J. Li, and H. Wu. “Wireless backhaul of dense small cell

networks with high dimension MIMO”. In Proceedings in IEEE Globecom Work-

shops, December, 2014.

[32] B. Luo, Q. Cui, X. Tao, and A. Dowhuszko. “On the optimal power allocation

for coordinated wireless backhaul in OFDM-Based relay systems”. In IEEE

International Conference on Commununication (ICC), pages 5625–5629, June, 2013.

[33] Ö. Bulakci, A. S. Nedelcu, A. B. Salehand S. Redana, and J. Hämäläinen. “Im-

pact of Backhaul Subframe Misalignment on Uplink System Performance of

LTE-Advanced Relay Networks”. In IEEE Vehicular Technology Conference (VTC

Fall), September, 2012.

[34] B.B. Haile, E. Mutafungwa, and J. Hämäläinen. “Use of coordinated multipoint

transmission for relaxation of relay link bottlenecks”. In 79th IEEE Vehicular

Technology Conference (VTC), pages 1–5, May, 2014.

[35] B.B. Haile, E. Mutafungwa, and J. Hämäläinen. “Coordinated multi-point

transmission for relaxation of self-backhauling bottlenecks in heterogeneous

networks”. EURASIP Journal on Wireless Communication and Networking,

March, 2015.

[36] J-M. Conrat, Q. H. Chu, I. Maaz, and J-C Cousin. “Path loss model comparison

for LTE-Advanced relay backhaul link in urban environment”. In 8th European

Conference on Antennas and Propagation (EuCAP), April, 2014.

[37] I. Rodriguez, C. Coletti, and T. B. Sorensen. “Evaluation of Potential Relay

Locations in a Urban Macro-Cell Scenario with Applicability to LTE-A”. In

IEEE 75th Vehicular Technology Conference (VTC Spring), May, 2012.

107

References

[38] M. I. H. bin Mansor and H. A. M. Ramli. “Performance Study of Path Loss

Models for LTE-A Relay Stations”. In 2016 International Conference on Computer

and Communication Engineering (ICCCE), July, 2016.

[39] M. Hamid and I. Kostanic. “Path loss models for LTE and LTE-A Relay Sta-

tions”. Universal Journal of Communications and Network, 1:119–126, 2013.

[40] J. Gan, Z. Guo, K Sandlund, J. Liu, X. Shen, R. Fan, W. Liu, H. Wang, and G. Liu.

“LTE In-Band Relay Prototype and Field Measurement”. In IEEE 75th Vehicular

Technology Conference (VTC Spring), pages 1–5, Yokohama, May, 2012. IEEE.

[41] Y. Qian, Z. Guo, R. Fan, H. Wang, J. Liu, Y. Yan, X. Shen, and Z. Hu. “Improv-

ing outdoor to indoor coverage by use of TD-LTE in-band relay ”. In IEEE 24th

Annual International Symposium on Personal, Indoor, and Mobile Radio Communi-

cations (PIMRC), pages 2658–2662, London, September, 2013. IEEE.

[42] A. B. Saleh, S. Redana, J. Hämäläinen, and B. Raaf. “On the coverage extension

and capacity enhancement of inband relay deployments in LTE-advanced net-

works”. Journal of Electrical and Computer Engineering - Special issue on LTE/LTE-

Advanced cellular communication networks, page 12, May, 2010.

[43] A. B. Saleh, Ö. Bulakci, S. Redana, B. Raaf, and J. Hämäläinen. “Enhancing

LTE-advanced relay deployments via biasing in cell selection and handover

decision”. In IEEE 21st International Symposium on Personal Indoor and Mobile

Radio Communications (PIMRC), pages 2277–2281, September, 2010.

[44] A. B. Saleh, S. Redana, B. Raaf, and J. Hämäläinen. “Comparison of relay and

pico eNB deployments in LTE-advanced ”. In 70th IEEE Vehicular Technology

Conference (VTC), pages 1–5, September September, 2009.

[45] T. Beniero, S. Redana, J. Hämäläinen, and B. Raaf. “Effect of relaying on cov-

erage in 3GPP LTE-Advanced”. In IEEE 69th Vehicular Technology Conference,

pages 1–5. IEEE, April April, 2009.

[46] R. Schoenen, W. Zirwas, and B. H. Walke. “Capacity and Coverage Analysis of

a 3GPP-LTE Multihop Deployment Scenario”. In IEEE International Conference

on Communications (ICC) Workshops, May, 2008.

[47] S. Nagata, Y. Yan, X. Gao, A. Li, H. Kayama, T. Abe, and T. Nakamura. “Inves-

tigation on system performance of L1/L3 relays in LTE-advanced downlink”.

In IEEE 73rd Vehicular Technology Conference, May, 2011.

[48] M. Minelli, M. Coupechoux, J-M. Kelif, M. Ma, and P. Godlewski. “Relays-

enhanced LTE-Advanced networks performance studies”. In 34th IEEE Sarnoff

Symposium, May 2011.

[49] O. N. C. Yilmaz, E. Mutafungwa, and J. Hämäläinen. “Performance of relay en-

hanced LTE-Advanced networks for selected suburban scenarios in emerging

market environments”. In International Symposium on Wireless Communication

Systems (ISWCS), August, 2012.

[50] Z. Ren, A. B. Saleh, Ö. Bulakci, S. Redana, B. Raaf, and J. Hämäläinen.

“Joint interference coordination and relay cell expansion in LTE-Advanced net-

works”. In IEEE Wireless Communications and Networking Conference (WCNC),

April, 2012.

108

References

[51] A. Yaver, D. Kolmas, and J. Lachowski. “Resource utilization with relays in LTE-

Advanced networks”. In 24th Canadian Conference on Electrical and Computer

Engineering (CCECE), May, 2011.

[52] Z. Ma, Y. Zhang, K. Zheng, W. Wang, and M. Wu. “Performance of 3GPP LTE-

Advanced networks with Type I relay nodes”. In 5th International ICST Confer-

ence on Communications and Networking in China (CHINACOM), August, 2010.

[53] Ö. Bulakci, S. Redana, B. Raaf, and J. Hämäläinen. “System Optimization in

Relay Enhanced LTE-Advanced Networks via Uplink Power Control”. In IEEE

71st Vehicular Technology Conference (VTC 2010-Spring), May, 2010.

[54] A. Karaer, Ö. Bulakci, S. Redana, B. Raaf, and J. Hämäläinen. “Uplink per-

formance optimization in relay enhanced LTE-Advanced networks”. In IEEE

20th International Symposium on Personal, Indoor and Mobile Radio Communica-

tions, September, 2009.

[55] T. M. de Moraes, A. B. Saleh, G. Bauch, and E. Seidel. “QoS-Aware Traffic

Scheduling in LTE-Advanced Relay-Enhanced Networks”. In IEEE 77th Vehic-

ular Technology Conference (VTC Spring), June, 2013.

[56] Y. Wang, G. Feng, and Y. Zhang. “Cost-Efficient Deployment of Relays for LTE-

Advanced Cellular Networks”. In IEEE International Conference on Communica-

tions (ICC), June, 2011.

[57] F. Parzysz, M. Vu, and F. Gagnon. “Trade-Offs on Energy-Efficient Relay De-

ployment in Cellular Networks,”. In IEEE 80th Vehicular Technology Conference

(VTC Fall), September, 2014.

[58] X. Cheng and H. Yin. “Uplink opportunistic spectrum sharing in LTE-A system

with multiple indoor relays”. In 8th International ICST Conference on Communi-

cations and Networking in China (CHINACOM), August, 2013.

[59] R. Fantini, D. Sabella, and M. Caretti. “Energy Efficiency in LTE-Advanced Net-

works with Relay Nodes,”. In IEEE 73rd Vehicular Technology Conference (VTC

Spring), July, 2011.

[60] R. Fantini, D. Sabella, and M. Caretti. “An E3F based assessment of energy

efficiency of relay nodes in LTE-advanced networks”. In IEEE 22nd Interna-

tional Symposium on Personal Indoor and Mobile Radio Communications (PIMRC),

September, 2011.

[61] A. B. Saleh, Ö. Bulakci, S. Redana, B. Raaf, and J. Hämäläinen. “Evaluating the

energy efficiency of LTE-Advanced relay and Picocell deployments”. In IEEE

Wireless Communications and Networking Conference (WCNC), April, 2012.

[62] G. Dimić, D. Bajić, and M. Beko. “Relay Type 1a in LTE-Advanced: Can it

increase energy efficiency?,”. In URSI General Assembly and Scientific Symposium

(URSI GASS), August, 2014.

[63] H. Y. Lateef, C. F. Chiasserini, T. ElBatt, A. Mohamed, and M. Guizani. “To-

wards energy efficient relay placement and load balancing in future wireless

networks”. In IEEE 25th Annual International Symposium on Personal, Indoor, and

Mobile Radio Communication (PIMRC), September, 2014.

109

References

[64] L. Mroueh and E. Vivier. “Energy efficient relaying PHY-MAC strategy for

LTE-advanced networks”. In International Symposium on Wireless Communica-

tion Systems (ISWCS), August, 2012.

[65] J.-J. Chen, J.-M. Liang, and Z.-Y. Chen. “Energy-efficient uplink radio resource

management in LTE-advanced relay networks for Internet of Things”. In In-

ternational Wireless Communications and Mobile Computing Conference (IWCMC),

August, 2014.

[66] D. Sabella, M. Caretti, and R. Fantini. “Energy saving schemes for self-

backhauled Small Cells in LTE-Advanced networks”. In IEEE Wireless Com-

munications and Networking Conference Workshops (WCNCW), April, 2014.

[67] E. Lang, S. Redana, and B. Raaf. “Business impact of relay deployment for

coverage extension in 3GPP LTE-advanced”. In IEEE International Conference

on Communications Workshops (ICC), pages 1–5, June, 2009.

[68] A. B. Saleh, S. Redana, B. Raaf, T. Riihonen, J. Hämäläinen, and R. Wichman.

“Performance of amplify-and-forward and decode-and-forward relays in LTE-

Advanced”. In 70th IEEE Vehicular Technology Conference, pages 1–5, Septem-

ber, 2009.

[69] E. Onu and O. Alani. “Collaborative algorithm for resource allocation in LTE-

Advanced relay networks”. In 9th International Symposium on Communication

Systems, Networks & Digital Signal Processing (CSNDSP), July, 2014.

[70] Ö. Bulakci, A. B. Saleh, Z. Ren, S. Redana, B. Raaf, and J. Hämäläinen. “Two-

step resource sharing and uplink power control optimization in LTE-Advanced

relay networks”. In 8th International Workshop on Multi-Carrier Systems & Solu-

tions (MC-SS), May, 2011.

[71] Ö. Bulakci, S. Redana, B. Raaf, and J. Hämäläinen. “Impact of power control

optimization on the system performance of relay based LTE-Advanced hetero-

geneous networks”. Journal of Communications and Networks, 13:345–359, Febru-

ary, 2012.

[72] C. Hoymann, W. Chen, J. Montojo, A. Golitschek, C. Koutsimanis, and X. Shen.

“Relaying operation in 3GPP LTE: challenges and solutions”. IEEE Communi-

cations Magazine, 50, February, 2012.

[73] A. BenMimoune, F. A. Khasawneh, M. Kadoch, and B. Rong. “Resource Allo-

cation Framework in 5G Multi-Hop Relay System”. In IEEE Global Communica-

tions Conference (GLOBECOM), December, 2015.

[74] Z. Zhang, K. Long, A. V. Vasilakos, and L. Hanzo. “Full-Duplex Wireless Com-

munications: Challenges, Solutions, and Future Research Directions”. Proceed-

ings of the IEEE, 104:1369–1409, February, 2016.

[75] K h. Liu and P. Lin. “Toward self-sustainable cooperative relays: state of the art

and the future”. IEEE Communications Magazine, 53:56 – 62, June, 2015.

[76] Z. Wei, X. Zhu, S. Sun, Y. Huang, A. Al-Tahmeesschi, and Y. Jiang. “Energy-

Efficiency of Millimeter-Wave Full-Duplex Relaying Systems: Challenges and

Solutions”. IEEE Access: Green Communications and Networking for 5G Wireless,

4:4848 – 4860, July, 2016.

110

References

[77] E. Basar. “Index modulation techniques for 5G wireless networks”. IEEE Com-

munications Magazine, 54:168 – 175, July, 2016.

[78] N. Nomikos, T. Charalambous, I. Krikidis, D. N. Skoutas, D. Vouyioukas, M. Jo-

hansson, and C. Skianis. “A Survey on Buffer-Aided Relay Selection”. IEEE

Communications Surveys & Tutorials, 18:1073 – 1097, December, 2015.

[79] A. Papadogiannis, M. Farber, A. Saadani, M. D. Nisar, P. Weitkemper, T. M.

de Moraes, J. Gora, N. Cassiau, D. Ktenas, J. Vihriala, M. Khanfouci, and

T. Svensson. “Pass It on: Advanced Relaying Concepts and Challenges for Net-

works Beyond 4G”. IEEE Vehicular Technology Magazine, 9:29 – 37, May, 2014.

[80] O. Delgado and F. Labeau. “D2D Relay Selection and Fairness on 5G Wireless

Networks”. In IEEE Globecom Workshops (GC Wkshps), December, 2016.

[81] P. V. Mekikis, E. Kartsakli, L. Alonso, and C. Verikoukis. “Flexible aerial re-

lay nodes for communication recovery and D2D relaying”. In IEEE 5th Global

Conference on Consumer Electronics. IEEE, October, 2016.

[82] S. Biswas, S. Vuppala, J. Xue, and T. Ratnarajah. “On the Performance of Re-

lay Aided Millimeter Wave Networks”. IEEE Journal of Selected Topics in Signal

Processing, December, 2015.

[83] G. Zheng, C. Hua, R. Zheng, and Q. Wang. “Toward Robust Relay Placement in

60 GHz mmWave Wireless Personal Area Networks with Directional Antenna”.

IEEE Transactions on Mobile Computing, 15:762 – 773.

[84] N. Zlatanov, R. Schober, and P. Popovski. “Buffer-Aided Relaying with Adap-

tive Link Selection”. IEEE Journal on Selected Areas in Communications, 31:1530

– 1542, August, 2013.

[85] B. F-Boroujeny. “OFDM Versus Filter Bank Multicarrier”. IEEE Signal Process-

ing Magazine, 28:92 – 112, April, 2011.

[86] M. Shaat and F. Bader. “Comparison of OFDM and FBMC performance in

multi-relay cognitive radio network”. In International Symposium on Wireless

Communication Systems (ISWCS), August, 2012.

[87] D. Gregoratti and X. Mestre. “AF relaying for FBMC signals”. In European

Conference on Networks and Communications (EuCNC), June,2014.

[88] M. Woltering, M. Zhang, D. Wuebben, and A. Dekorsy. “Comparison of Gen-

eral Multi-Carrier Schemes in Two Way Relaying Channels”. In Proceedings of

the 20th International ITG Workshop on Smart Antennas (WSA), March, 2016.

[89] B. Panzner, V. Pauli, L. Yu, and I. Viering. “System level modeling of in-band

wireless backhaul for 5G mmW”. In International Symposium on Wireless Com-

munication Systems (ISWCS), August, 2015.

[90] Z. Lan, J. Wang, J. Gao, C-S. Sum, F. Kojima, T. Baykas, H. Harada, and

S. Kato. “Directional Relay with Spatial Time Slot Scheduling for mmWave

WPAN Systems”. In IEEE 71st Vehicular Technology Conference (VTC 2010-

Spring), May, 2010.

[91] S. Rangan, T. S. Rappaport, and E. Erkip. “Millimeter-Wave Cellular Wireless

Networks: Potentials and Challenges”. Proceedings of the IEEE, 102:366–385,

February, 2014.

111

References

[92] I. Krikidis and G. Zheng. Advanced Relay Technologies in Next Generation Wireless

Communications. Institution of Engineering and Technology, 2016.

[93] M. Li, M. Lin, W-P. Zhu, Y. Huang, K-K. Wong, and Q. Yu. “Performance Anal-

ysis of Dual-Hop MIMO AF Relaying Network With Multiple Interferences”.

IEEE Transactions on Vehicular Technology, 66:1891–1897, May, 2016.

[94] M. R. Bhatnagar and Arti M.K. “Selection Beamforming and Combining in

Decode-and-Forward MIMO Relay Networks”. IEEE Communications Letters,

17:1556–1559, July,2013.

[95] J. Chen, X. Chen, T. Liu, and L. Lei. “Toward Green and Secure Communi-

cations over Massive MIMO Relay Networks: Joint Source and Relay Power

Allocation”. IEEE Access, 5:869–880, January, 2017.

[96] H. Shen, W. Xu, and C. Zhao. “Transceiver Optimization for Full-Duplex Mas-

sive MIMO AF Relaying With Direct Link”. IEEE Access, 4:8857 – 8864, Decem-

ber, 2016.

[97] J. Chen, X. Chen, W. H. Gerstacker, and D. W. K. Ng. “Resource Allocation for

a Massive MIMO Relay Aided Secure Communication”. IEEE Transactions on

Information Forensics and Security, 11:1700 – 1711, April, 2016.

[98] C. D. Ho, H. Q. Ngo, M. Matthaiou, and T. Q. Duong. “Multi-way massive

MIMO relay networks with maximum-ratio processing”. In International Con-

ference on Recent Advances in Signal Processing, Telecommunications & Computing

(SigTelCom), February, 2017.

[99] D. Kudathanthirige and G. A. Baduge. “Multi-Cell Multi-Way Massive MIMO

Relay Networks”. IEEE Transactions on Vehicular Technology, PP, January, 2017.

[100] 3rd Generation Partnership Project; Technical Specification Group Radio Ac-

cess Network; “Evolved Universal Terrestrial Radio Access (E-UTRA); Physi-

cal layer for relaying operation (Release 14),” 3GPP Tech. Spec. 36.216 v14.0.0,

April, 2017.

[101] E. Hossain et al. Cooperative Cellular Wireless Networks. Cambridge University

Press, UK, 1st edition, 2011.

[102] 3rd Generation Partnership Project; Technical Specification Group (TSG) Radio

Access Network (RAN), Workng Group 1 (WG1) “Application of network cod-

ing in LTE-advanced relay,” 3GPP Rep. R1-082327, Samsung, Poland, July, 2008.

[103] S.-Y. R. Li, R.-W. Yeung, and N. Cai. “Linear Network Coding”. IEEE Transac-

tions on Information Theory, 49:371–381, February, 2003.

[104] Z. Ma, Z. Zhang, Z. Ding, P. Fan, and H. Li. “Key techniques for 5G wireless

communications: network architecture, physical layer, and MAC layer perspec-

tives”. Science China Information Sciences, 58:1–20, February, 2015.

[105] A. Osseiran, K. Doppler, C. Ribeiro, M. Xiao, M. Skoglund, and J. Manssour.

“Advances in device-to-device communications and network coding for IMT-

advanced”. In In proceedings of ICT-Mobile Summit Conference, 2009.

[106] B. Nazer and M. Gastpar. “Reliable Physical Layer Network Coding”. Proceed-

ings of the IEEE, 99:438–460, January, 2011.

112

References

[107] S. Shukla, V. T. Muralidharan, and B. S. Rajan. “Wireless Network Coded Ac-

cumulate Compute and Forward Two Way Relaying”. IEEE Transactions on Ve-

hicular Technology, 65:1367 – 1381, March, 2015.

[108] J. Yue, Z. Lin, B. Vucetic, G. Mao, M. Xiao, B. Bai, and K. Pang. “Network

Code Division Multiplexing for Wireless Relay Networks”. IEEE Transactions

on Wireless Communications, 14:5736–5749, June, 2015.

[109] Q-T. Vien, T. A. Le, H. X. Nguyen, and H. Phan. “A Secure Network Cod-

ing Based Modify-and-Forward Scheme for Cooperative Wireless Relay Net-

works”. In IEEE 83rd Vehicular Technology Conference (VTC Spring), May, 2016.

[110] N. T. Do, V. Nguyen, Q. Bao, and B. An. “A relay selection protocol for wire-

less energy harvesting relay networks”. In International Conference on Advanced

Technologies for Communications (ATC), October, 2015.

[111] S. Vahidian, S. Aïssa, and S. Hatamnia. “Relay Selection for Security-

Constrained Cooperative communication in the Presence of Eavesdropper’s

Overhearing and Interference”. IEEE Wireless Communications Letters, 4:577–

580, December, 2015.

[112] K. R. Reddy and A. Rajesh. “Best relay selection using co-operative game the-

ory: MANETs ”. In International Conference on Communication and Signal Pro-

cessing (ICCSP),. IEEE, April, 2016.

[113] G. Al Sukkar, Z. A. Shafeeq, and A. Al Amayreh. “Best relay selection in a

multi-relay nodes system under the concept of cognitive radio”. In Interna-

tional Conference on Pervasive and Embedded Computing and Communication Sys-

tems (PECCS). IEEE, June, 2015.

[114] G. Xie, Y. LiuJinChun, and G. Li. “Sort-based relay selection algorithm for

decode-and-forward relay system”. Science China Information Sciences, 56:1–8,

June, 2013.

[115] M. Seyfi, S. Muhaidat, and J. Liang. “Performance Analysis of Relay Selection

With Feedback Delay and Channel Estimation Errors”. IEEE Signal Processing

Letters, 18:67–70, January, 2011.

[116] P. Liu, C. Nie, E. Erkip, and S. Panwar. “Robust Cooperative Relaying in a

Wireless LAN: Cross-Layer Design and Performance Analysis”. In IEEE Global

Telecommunications Conference (GLOBECOM), December, 2009.

[117] Q. Zhang, J. Jia, and J. Zhang. “Cooperative relay to improve diversity in

cognitive radio networks”. IEEE Communications Magazine, 47:111–117, Febru-

ary, 2009.

[118] L. Zhai, H. Ji, X. Li, and Y. Tang. “Coalition Graph Game for joint relay selec-

tion and resource allocation in cooperative cognitive radio networks”. In IEEE

Global Communications Conference (GLOBECOM), December, 2012.

[119] H. Zhu and G. Cao. “On improving the performance of IEEE 802.11 with mul-

tihop concepts”. In 12th International Conference on Computer Communications

and Networks, October, 2003.

[120] Z. Ding, H. Dai, and H. V. Poor. “Relay Selection for Cooperative NOMA”. IEEE

Wireless Communications Letters, 5:416 – 419, June, 2016.

113

References

[121] G. Ozcan and M. C. Gursoy. “Channel sensing and estimation in cognitive relay

networks”. In IEEE 13th International Workshop on Signal Processing Advances in

Wireless Communications (SPAWC), June, 2012.

[122] M. Salem, A. Adinoyi, M. Rahman, H. Yanikomeroglu, D. Falconer, Y-D. Kim,

E. Kim, and Y-C Cheon. “An Overview of Radio Resource Management in

Relay-Enhanced OFDMA-Based Networks”. IEEE Communications Surveys &

Tutorials, 12:422–438, April, 2010.

[123] Y. Saito, A. Benjebbour, A. Li, K. Takeda, Y. Kishiyama, and T. Nakamura.

lq‘System-level evaluation of downlink non-orthogonal multiple access

(NOMA) for non-full buffer traffic model”. In IEEE Conference on Standards for

Communications and Networking (CSCN), October, 2015.

[124] H. Holma and A. Toskala. LTE-Advanced: 3GPP Solution for IMT-Advanced. Wi-

ley and Sons Ltd, August 2012.

[125] J. Gora and S. Redana. “In-Band and out-band relaying configurations for dual-

carrier LTE-Advanced system”. In IEEE 22nd International Symposium on Per-

sonal Indoor and Mobile Radio Communications (PIMRC)., September, 2011.

[126] G. Liu, F. R. Yu, H. Ji, V. C. M. Leung, and X. Li. “In-Band Full-Duplex Relaying:

A Survey, Research Issues and Challenges”. IEEE Communications Surveys &

Tutorials, 17:500–524, January, 2015.

[127] T. Riihonen, S. Werner, and R. Wichman. “Spatial loop interference suppres-

sion in full-duplex MIMO relays”. In IEEE Conference on Signals, Systems and

Computers, 2009 Conference Record of the Forty-Third Asilomar, November, 2009.

[128] T. Riihonen, S. Werner, and R. Wichman. “Mitigation of Loopback Self-

Interference in Full-Duplex MIMO Relays”. IEEE Transactions on Signal Pro-

cessing, 59:5983 – 5993, August, 2011.

[129] J. N. Laneman, D.N.C. Tse, and G.W. Wornell. “Cooperative diversity in wire-

less networks: Efficient protocols and outage behavior”. IEEE Transactions on

Infomation Theory, 50:3062–3080, November, 2004.

[130] M. O. Hasna and M. S. Alouini. “Harmonic mean and end-to-end performance

of transmission systems with relays”. IEEE Transactions on Communications,

52:130–135, March, 2004.

[131] M. O. Hasna and M. S. Alouini. “End-to-end performance of transmission sys-

tems with relays over Rayleigh-fading channels”. in IEEE Transactions on Wire-

less Communications, 2:1126–1131, November, 2003.

[132] M. O. Hasna and M. S. Alouini. “A performance study of dual-hop transmis-

sions with fixed gain relays”. IEEE Transactions on Wireless Communications,

3:1963–1968, November, 2004.

[133] G. Kramer, M. Gastpar, and P. Gupta. “Cooperative strategies and capacity

theorems for relay networks”. IEEE Transactions on Infomation Theory, 51:3037–

3063, September August, 2005.

[134] J. Zheng, P. Sartori, and B. Wei. “Performance Analysis of Layer 1 Relays”. In

IEEE International Conference on Communications Workshops, June, 2009.

114

References

[135] D. Soldani and S. Dixit. “Wireless relays for broadband access [radio commu-

nications series]”. IEEE Communications Magazine, 46, March, 2008.

[136] G. Shen, K. Zhang, D. Wang, J. Liu, X. Leng, W. Wang, and S. Jin. “Multi-hop

relay operation modes”. Technical report, Alcatel Shanghai Bell, IEEE 802.16

Broadband Wireless Access Working Group, October, 2008.

[137] K. J. R. Jiu, A. K. Sadek, W. Su, and A. Kwasinski. Cooperative Communications

and Networking. Cambridge University Press, 2009.

[138] Y. W. P. Hong, W. J. Huang, and C. C J. Kuo. Cooperative and Communication and

Networking, Technologies and System Design. Springer New York, 2010.

[139] J. Sydir et al. “IEEE 802.16 Broadband Wireless Access Working Group:

Harmonized contribution on 802.16j (Mobile Multihop Relay) usage models”,

September, 2006.

[140] Ö. Bulakci, Z. Ren, C. Zhou, J. Eichinger, P. Fertl, and S. Stanczak. “Dynamic

Nomadic Node Selection for Performance Enhancement in Composite Fad-

ing/Shadowing Environments,”. In IEEE 79th Vehicular Technology Conference

(VTC), May, 2014.

[141] Ö. Bulakci, Z. Ren, C. Zhou, J. Eichinger, P. Fertl, D. G. Serrano, and S. Stanczak.

“Towards flexible network deployment in 5G: Nomadic node enhancement to

heterogeneous networks,”. In IEEE International Conference on Communication

Workshop (ICCW), June, 2015.

[142] A. Damnjanovic, J. Montojo, Y. Wei, T. Ji, T. Luo, M. Vajapeyam, T. Yoo, O. Song,

and D. Malladi. “A survey on 3GPP heterogeneous networks”. IEEE Wireless

Communications, 18, June, 2011.

[143] Z. Li, M. Moisio, M. A. Uusitalo, P. Lundén, C. Wijting, F. S. Moya, A. Yaver, and

V. Venkatasubramanian. “Overview on initial METIS D2D concept”. In 1st In-

ternational Conference on 5G for Ubiquitous Connectivity (5GU), November, 2014.

[144] G. Fodor, S. Parkvall, S. Sorrentino, P. Wallentin, Q. Lu, and N. Brahmi. “Device-

to-Device Communications for National Security and Public Safety”. 5G Wire-

less Technologies: Perspectives of the Next Generation Mobile Communications and

Networking, 2:1510–1520, December, 2014.

[145] G. Huang, J. Hu, S. Pan, Y. Zhang, H. Han, and G. Zhang. “D2D relaying based

multicast service in Public Safety Networks”. In 35th Chinese Control Conference

(CCC). IEEE, July, 2016.

[146] J. Deng, A. A. Dowhuszko, R. Freij, and Olav Tirkkonen. “Relay Selection and

Resource Allocation for D2D-Relaying under Uplink Cellular Power Control”.

In IEEE Globecom Workshops. IEEE, December, 2015.

[147] Ö. Bulakci, A. Prasad, J. Belschner, M. Ericson, I. Karls, H. Celik, M. Tesanovic,

R. Fantini, L. M Campoy, E. Pateromichelakis, F. S. Moya, G. Zimmermann, and

I. D. Silva. “Agile resource management for 5G: A METIS-II perspective”. In

IEEE Conference on Standards for Communications and Networking (CSCN), Octo-

ber, 2015.

[148] M. Fallgren and J. F. Monserrat. “Intermediate system evaluation results, ICT-

317669-METIS/D6.3”, August, 2014.

115

References

[149] J. F. Monserrat and M. Fallgren. “Report on simulation results and evaluations,

ICT-317669-METIS/D6.5 ”, March, 2015.

[150] H. Tullberg, P. Popovski, Z. Li, M. A. Uusitalo, A. Hoglund, Ö. Bulakci, M. Fall-

gren, and J. F. Monserrat. “The METIS 5G System Concept: Meeting the 5G

Requirements”. IEEE Communications Magazine, 54:132–139, December, 2016.

[151] Z. Ren, S. Stanczak, P. Fertl, and F. Penna. “Energy-aware activation of nomadic

relays for performance enhancement in cellular networks, ”. In IEEE Interna-

tional conference on communications, June June, 2014.

[152] H. Tullberg and M. Fallgren. “Final report on the METIS 5G system concept

and technology roadmap, ICT-317669-METIS/D6.6 ”, May, 2015.

[153] H. Tullberg, Z. Li, A. Hoglund, P. Fertl, D. G. Serrano, K. Pawlak, P. Popovski,

G. Mange, and Ö. Bulakci. “Towards the METIS 5G concept: First view on

Horizontal Topics concepts”. In European Conference on Networks and Communi-

cations (EuCNC), June, 2014.

[154] A. Jaziri, R. Nasri, and T. Chahed. “Congestion mitigation in 5G networks us-

ing drone relays”. In IEEE International on Wireless Communications and Mobile

Computing Conference (IWCMC), September, 2016.

[155] M. Krol, Y. Ji, S. Yamada, C. Borcea, L. Zhong, and K. Takano. “Extending

Network Coverage by Using Static and Mobile Relays during Natural Disas-

ters,”. In 30th International Conference on Advanced Information Networking and

Applications Workshops (WAINA), March, 2016.

[156] Z. Ren, S. Stanczak, and P. Fertl. “Activation of nomadic relay nodes in dynamic

interference environment for energy saving,”. In IEEE Global Communications

Conference (GLOBECOM), December, 2014.

[157] M. F. Feteiha and H. S. Hassanein. “Enabling Cooperative Relaying VANET

Clouds Over LTE-A Networks”. IEEE Transactions on Vehicular Technology,

64:1468 – 1479, April, 2015.

[158] K. Miranda, A. Molinaro, and T. Razafindralambo. “A survey on rapidly de-

ployable solutions for post-disaster networks,”. IEEE Communications Magazine,

54:117–123, April, 2016.

[159] J. Q. Bao and W. C. Lee. “Rapid Deployment of Wireless Ad Hoc Backbone

Networks for Public Safety Incident Management,”. In IEEE Global Telecommu-

nications Conference (GLOBECOM), November, 2007.

[160] H. Li, D. Branscomb, A Johnson, M. Baginski, L. Riggs, and G. Thomas View All

Authors. “Rapid Deployment Wireless Text Messaging Network for Disaster

Relief,”. In 7th International Conference on Wireless Communications, Networking

and Mobile Computing (WiCOM), September, 2011.

[161] Universal Mobile Telecommunications System (UMTS); LTE; “Requirements

for Evolved UTRA (E-UTRA) and Evolved UTRAN (E-UTRAN) (Release

9),” 3GPP Tech. Rep. 25.913 v9.0.0, February, 2010.

[162] “LTE, Requirements for further advancements for Evolved Universal Terres-

trial Radio Access (E-UTRA) (Release 10), ” 3GPP Tech. Rep. 36.913 v10.0.0,

April, 2011.

116

References

[163] J. T. J. Penttinen. The LTE-Advanced Deployment Handbook: The Planning Guide-

lines for the Fourth Generation Networks. John Wiley and Sons, Ltd, Chichester,

UK., 2016.

[164] O. Teyeb, V. V. Phan, B. Raaf, and S. Redana. “Dynamic relaying in 3GPP LTE-

Advanced networks”. EURASIP Journal on Wireless Communications and Net-

working, July, 2009.

[165] 3rd Generation Partnership Project; Technical Specification Group Radio Ac-

cess Network; “Study on LTE-based V2X Services (Release 14), ” 3GPP Tech.

Rep. 36.885, June, 2016.

[166] H. Yanikomeroglu. “Fixed and Mobile Relaying Technologies for Cellular Net-

works. In 2nd Workshop application and services in Wireless Networks, pages 75–81,

July, 2002.

[167] 3rd Generation Partnership Project; Technical Specification Group Radio Ac-

cess Network; “Full duplex configuration of Un and Uu subframes for Type I

relay, ” 3GPP Rep. R1-100139, Valencia, Spain, January, 2010.

[168] 3rd Generation Partnership Project; Technical Specification Group Radio Ac-

cess Network; “Text proposal on inband full duplex relay for TR 36.814, ” 3GPP

Rep. R1-101659, SA, USA, February, 2010.

[169] 3rd Generation Partnership Project; Technical Specification Group Radio Ac-

cess Network; “Evolved Universal Terrestrial Radio Access (E-UTRA); Relay ar-

chitectures for E-UTRA (LTE-Advanced) (Release 9), ” 3GPP Tech. Rep. 36.806

v9.0.0, March, 2010.

[170] “LTE; Evolved Universal Terrestrial Radio Access (E-UTRA) and Evolved Uni-

versal Terrestrial Radio Access Network (E-UTRAN); Overall description; Stage

2 (Release 10), ” 3GPP Tech. Spec. 36.300 v10.4.0, March, 2011.

[171] “LTE; Evolved Universal Terrestrial Radio Access Network (E-UTRAN); S1

Application Protocol (S1AP) (Release 14),” 3GPP Tech. Spec. 36.413 v14.2.0,

March, 2017.

[172] “LTE; Evolved Universal Terrestrial Radio Access Network (E-UTRAN); X2 Ap-

plication Protocol (X2AP) (Release 10), ” 3GPP Tech. Spec. 36.423 v10.4.0 , Jan-

uary, 2012.

[173] 3rd Generation Partnership Project Technical Specification Group Radio Ac-

cess Network; “Evolved Universal Terrestrial Radio Access (E-UTRA); Relay

radio transmission and reception (Release 11),” 3GPP Tech. Rep. 36.826 v11.0.0,

September, 2012.

[174] Texas Instruments, “On the design of relay node for LTE-advanced ”3GPP

Technical Specification Group Radio Access Network Working Group 1, Seoul,

Korea, Tech. Rep. R1-091294, March, 2009.

[175] S. Sesia, I. Toufik, and M. Baker. “LTE - The UMTS Long Term Evolution: From

Theory to Practice”. Wiley and Sons Ltd, UK, 2011.

[176] O. A. Elgendy, M. H. Ismail, and K. Elsayed. “Max-min fair resource allocation

for LTE-advanced relay-enhanced cells,”. In IEEE Wireless Communications and

Networking Conference (WCNC), April, 2014.

117

References

[177] A. Abdallah, D. Serhal, and K. Fakih. “Relaying Techniques for LTE-

Advanced,”. In Proceedings of European Wireless 2015; 21th European Wireless

Conference, July,2015.

[178] L. Ding, P. Wu, H. Wang, Z. Pan, and X. You. “Lifetime maximization routing

with network coding in wireless multihop networks”. Science China Information

Sciences, 56:1–15, June, 2013.

[179] A. Bletsas, A. Khisti, D.P. Reed, and A. Lippman. “A simple Cooperative diver-

sity method based on network path selection ”. IEEE Journal on Selected Areas

in Communications, 24(3):659–672, March, 2006.

[180] M. Sadeghi and A. M. Rabiei. “A Fair Relay Selection Scheme for a DF Cooper-

ative Network With Spatially Random Relays ”. IEEE Transactions on Vehicular

Technology, July, 2017.

[181] J. Deissner and G.P. Fettweis. “A study on hierarchical cellular structures with

inter-layer reuse in an enhanced GSM radio network”. IEEE International Work-

shop on Mobile Multimedia Communications (MoMuC), November, 1999.

[182] Y. Shi, M. Li, X. Xiong, G. Han, L. Wan, and X. Dong. “A flexible backhaul ar-

chitecture for LTE-Advanced”. In International Conference on Connected Vehicles

and Expo (ICCVE), November, 2014.

[183] M. F. Feteiha and H. S. Hassanein. “Decode-and-Forward vehicular relaying for

2x2 MIMO LTE-advanced downlink,”. In International Wireless Communications

and Mobile Computing Conference (IWCMC), October, 2015.

[184] M. Bhatnagar. “On the capacity of decode-and-forward relaying over Rician

fading channels”. IEEE Communications Letters, 17(6):1100–1103, May, 2013.

[185] Y. Zhang, F. Ke, S. Feng, and X. Zhu. “Tight bounds for ergodic capacity of

cooperative relaying over Rayleigh fading channels”. In International conference

on Computational Intelligence and Communication Networks (CICN), pages 313–

316. IEEE, November, 2014.

[186] J. Yang, Y. Wang, and Z. Xie. “Performance analysis of dual-hop DF relaying in

G fading channel in the presence of co-channel interference”. In International

Conference on Wireless Communications & Signal Processing (WCSP), pages 1–6.

IEEE, December, 2013.

[187] Y. Sun, X. Zhong, X. Chen, S. Zhou, and J. Wang. “Ergodic capacity of decode-

and-forward relay strategies over general fast fading channels”. Electronics Let-

ters, 47:148–150, January, 2011.

[188] G. Farhadi and N. C. Beaulieu. “On the ergodic capacity of multi-hop wireless

relaying systems”. IEEE Transactions on Wireless Communications, 8:2286–2291,

May, 2009.

[189] X. Qu, X. Xu, H. Li, X. Tao, and H. Tian. “Analysis of outage capacity for dual-

hop relay and optimal power allocation”. In IET International Conference on Wire-

less Sensor Network, pages 194–197. IEEE, November, 2010.

[190] W-G. Li and M. Chen. “Outage capacity of dual-hop decode-and-forward re-

laying system over generalized fading channels”. In IEEE International Confer-

ence on Future Computer and Communication (ICFCC), volume 3, pages V3–827–

V3831, May, 2010.

118

References

[191] M. A. Hankal, I. A. Eshrah, and H. M. Tawfik. “Performance of the dual-hop

decode and forward relaying systems over Nakagami-m fading channels”. In

31st National Radio Science Conference (NRSC), pages 219–227. IEEE, April, 2014.

[192] N. C. Beaulieu and J. Hu. “A closed-form expression for the outage probability

of decode-and-forward relaying in dissimilar Rayleigh fading channels”. IEEE

Communications Letters, 10:813–815, December, 2006.

[193] J. Hu and N. C. Beaulieu. “Closed-Form Expressions for the Outage and Error

Probabilities of Decode-and-Forward Relaying in Dissimilar Rayleigh Fading

Channels”. In IEEE International conference on Communication, pages 5553–5557,

June, 2007.

[194] B. R. Manoj, R. K. Mallik and M. R. Bhatnagar. “Buffer-Aided Multi-Hop DF

Cooperative Networks: A State-Clustering Based Approach ”. IEEE Transac-

tions on Communications, 64(12):4997–5010, December, 2016.

[195] N. Yi, Y. Ma, and R. Tafazolli. “Joint Rate Adaptation and Best-Relay Selec-

tion Using Limited Feedback”. IEEE Transactions on Wireless Communications,

12(6):2797–2805, June 2013.

[196] Y. Ma, R. Tafazolli, Y. Zhang, and C. Qian. “Adaptive Modulation for Oppor-

tunistic Decode-and-Forward Relaying”. IEEE Transactions on Wireless Commu-

nications, 10(7):2017–2022, July 2011.

[197] Z. Hasan, H. Boostanimehr, and V. K. Bhargava. “Green Cellular Networks: A

Survey, Some Research Issues and Challenges ”. IEEE Communications Surveys

& Tutorials, 13(4):524–540, August 2011.

[198] S. McLaughlin, P. M. Grant, J.S. Thompson, H. Haas, D.I. Laurenson, C. Khi-

rallah, Y. Hou, and R. Wang. “Techniques for improving cellular radio base

station energy efficiency ”. IEEE Wireless Communications, 18(5):10–17, August

2011.

[199] M. Liu, J. Zhang, and P. Zhang. “Outage probability of dual-hop multiple an-

tenna relay systems with interference at the relay and destination ”. Interna-

tional Journal of Antennas and Propagation, 2014.

[200] Q. Cui, Xianjun Yang, J. Hämäläinen, Xiaofeng Tao, and Ping Zhang. “Opti-

mal energy-efficient relay deployment for the bidirectional relay transmission

schemes ”. IEEE Transactions on Vehicular Technology, 63(6), July 2014.

[201] Ö. Bulakci, A. Bou Saleh, S. Redana, B. Raaf, and J. Hämäläinen. “Resource

Sharing in LTE Advanced Relay Networks: Uplink Optimization Methods and

Performance Analysis ”. European Transactions on Telecommunications, 24(1):32–

48, January 2013.

[202] D. Qin and Y. Wang. “Capacity analysis of two-hop multichannel relaying with

subchannel pairing ”. IEEE Communications Letters, 19(10):1846–1849, 2015.

[203] S. Wang, H. Peng, and X. Liu. “Sum-power minimization in multiuser single-

DF-relay Networks with Direct Links ”. International Journal of Distributed Sensor

Networks, 2016.

[204] M. R. Bhatnagar, R. K. Mallik, and O. Tirkkonen. “Performance evaluation of

best-path selection in a multihop decode-and-forward cooperative system ”.

IEEE Transactions on Vehicular Technology, 65(4):2722 – 2728, 2016.

119

References

[205] M. R. Bhatnagar. “Performance analysis of a path selection scheme in multi-

hop decode-and-forward protocol ”. IEEE Communications Letters, 16(12):1980–

1983, December, 2012.

[206] L. J. Greenstein, S. S. Ghassemzadeh, V. Erceg, and D. G. Michelson. Ricean

K-Factors in narrow-band fixed wireless channels: Theory, experiments, and

statistical models. IEEE Transactions on Vehicular Technology, 58(8):4000–4012,

Oct. 2009.

[207] M. K. Simon and M. S. Alouini. Digital communication over fading channels, vol-

ume 95. John Wiley & Sons, 2005.

[208] X. Liu. “Outage behavior of LTE-A with non-identical Rician relay links”. In

Proc. 2016 IEEE 83rd Vehicular Technology Conference (VTC Spring), pages 1–5,

May 2016.

[209] Andreas F Molisch. Wireless Communications. John Wiley & Sons, 2007.

[210] K. Loa, C. Wu, S. Sheu, Y. Yuan, M. Chion, D. Huo, and L. Xu. “IMT-Advanced

relay standards [WiMAX/LTE update]”. IEEE Communications Magazine, 48,

August, 2010.

[211] I. S. Gradshteyn, I.M. Ryzhik, and D. Zwillinger. Table of integrals, series and

products. Elsevier Academic Press, 2007.

[212] A. Papoulis. Probability, Random Variables, and Stochastic Processes. McGraw-Hill

Inc, 3rd edition, 1991.

[213] D. Tse and P. Viswanath. Fundamental of Wireless Communication. Cambridge

University Press, 2005.

[214] “Wolfram Research, Exponential Integral Functions” [Online]. Avail-

able: http://functions.wolfram.com/01.03.21.0128.01. [Last Accessed]:

September, 2016.

[215] Milton Abramowitz and Irene. A. Stegun. Handbook of Mathematical Functions.

National Bureau of Standards, Washington, DC, 1972.

[216] P. Bhat, S. Nagata, L. Campoy, I. Berberana, T. Derham, G. Liu, X. Shen, P. Zong,

and J. Yang. “LTE-Advanced: An operator perspective”. IEEE Communications

Magazine, 50:104–114, February, 2012.

[217] J. Y. Ryu and W. Choi. “Balanced linear precoding in decode-and-forward

based MIMO relay communications ”. IEEE Transactions on Wireless Commu-

nications, 10:2390–2400, May, 2011.

[218] Y. Lu, N. Yang, H. Dai, and X. Wang. “Opportunistic Decode-and-Forward

Relaying With Beamforming in Two-Wave With Diffuse Power Fading ”. IEEE

Transactions on Vehicular Technology, 61:3050–3060, June,2012.

[219] Z. F. Xu, P. Y. Fan, H. C. Yang, K. Xiong, M. Lei, and S. Yi. “Optimal beamform-

ing for MIMO decode-and-forward relay channels ”. In in Proceeding of IEEE

Global Communications Conference (GLOBECOM), December, 2012.

[220] R. Irmer and F. Diehm. “On coverage and capacity of relaying in LTE-advanced

in example deployments ”. In IEEE 19th International Symposium on Personal,

Indoor and Mobile Radio Communications, 2008. PIMRC, September, 2008.

120

References

[221] C. Coletti, P. Mogensen, and R. Irmer. “Performance Analysis of Relays in LTE

for a Realistic Suburban Deployment Scenario ”. In 73rd IEEE Vehicular Tech-

nology Conference (VTC Spring), May,2011.

[222] Z. Chen, H. Liu, and W. Wang. “Optimal transmit strategy of a two-hop

decode-and-forward MIMO relay system with mean and covariance feedback

”. IEEE Communications Letters, 14, June, 2010.

[223] K. Xiong, P. Fan, Z. Xu, H. C. Yang, and K. B. Letaief. “Optimal Cooperative

Beamforming Design for MIMO Decode-and-Forward Relay Channels ”. IEEE

Transactions on Signal Processing, 62(6):1476–1489, January, 2014.

[224] M. Nakagami. “The m-distribution - a general formula of intensity distribution

of rapid fading,”. Proceedings of a Symposium held in Statistical Methods in Radio

Wave Propagation, June, 1958.

[225] W. Braun and U. Dersch. “A physical mobile radio channel model ”. IEEE

Transactions on Vehicular Technology, 40(2):472–482„ May 1991.

[226] M. D. Yacoub. “The κ−μ distribution and the η−μ distribution ”. IEEE Antenna

and Propagation Magazine, 49:68–81, February, 2007.

[227] M. D. Yacoub. “The α− μ Distribution: A Physical Fading Model for the Stacy

Distribution ”. IEEE Transactions on Vehicular Technology ( Volume: 56, Issue: 1,,

56:27–34, January, 2007.

[228] M. Husso, J. Hämäläinen, R. Jäntti, J. Li, E. Mutafungwa, R. Wichman,

Z. Zheng, and A.M. Wyglinski. “Interference Mitigation by Practical Transmit

Beamforming Methods in Closed Femtocells ”. in EURASIP Journal on Wireless

Communications and Networking, Special Issue on Femtocell Networks, 2010.

[229] A. Dowhuszko, M. Husso, and J. Hämäläinen. “Combined Transmit Beam-

forming and Channel-Aware Scheduling for Interference Mitigation in Femto-

cells,”. in EURASIP Journal on Wireless Communications and Networking, 2012.

[230] H.C. Lee, D.C. Oh, and Y.H. Lee. “Coordinated user scheduling with transmit

beamforming in the presence of inter-femtocell interference ”. In IEEE Interna-

tional Conference on Communications (ICC), June, 2011.

[231] S. Park, W. Seo, S. Choi, and D. Hong. “A beamforming codebook restriction

for cross-tier interference coordination in two-tier femtocell networks ”. IEEE

Transactions on Vehicular Technology, 60:1651–1663, May, 2011.

[232] A. Dowhuszko and J. Hämäläinen. “Performance of Transmit Beamforming

Codebooks with Separate Amplitude and Phase Quantizations ”. IEEE Signal

Processing Letters, 22:813–817, July, 2015.

[233] Hämäläinen J., R. Wichman, A. A. Dowhuszko, and G. Corral-Briones. “Capac-

ity of generalized UTRA FDD closed-loop transmit diversity modes”. Wireless

Personal Communication, 54:467–484, August, 2010.

[234] D. Morales-Jimenez, F. J. Lopez-Martinez, E. Martos-Naya, J. F. Paris, and

A. Lozano. “Connections Between the Generalized Marcum Q-Function and a

Class of Hypergeometric Functions ”. IEEE Transactions on Information Theory,

60:1077 – 1082, February, 2014.

121

References

[235] S. Aguirre, L.H. Loew, and Y. Lo. “Radio propagation into buildings at 912,

1920, and 5990 MHz using microcells,”. In Third Annual International Conference

on Universal Personal Communications, pages 129–134. IEEE, October, 1994.

[236] European Commission. “COST Action 231 Digital mobile radio towards future

generation systems, Final report, Technical Report European Communities”,

1999.

[237] H. Holma and A. Toskala. WCDMA for UMTS: HSPA Evolution and LTE. John

Wiley and Sons Ltd, Chichester, UK, 5th edition, August, 2010.

[238] Ericsson, “Optimizing the Indoor Experience,”White Paper, 2013, [Online].

Available: http://www.ericsson.com/res/docs/2013/real-performance-

indoors.pdf.

[239] Ericsson regional report Europe, “The Indoor Influence,” April, 2015, [Online].

Available:https://www.ericsson.com/res/docs/2015/consumerlab/ericsson-

consumerlab-the-indoor-influence-europe.pdf.

[240] M. Paolini. “Beyond data caps. An analysis of the uneven

growth in data traffic,” Senza Fili Consulting, . [Online]. Avail-

able: http://www.senzafiliconsulting.com. Technical report, February, 2011.

[241] Ericsson, “Mobility Report on the pulse of

the network society,” November, 2016, [Online].

Available:https://www.ericsson.com/assets/local/mobility-

report/documents/2016/ericsson-mobility-report-november-2016.pdf.

[242] E. Semaan, F. Harrysson, A. Furuskär, and H. Asplund. “Outdoor-to-Indoor

coverage in high frequency bands”. In IEEE Globecom Workshops, Decem-

ber, 2014.

[243] C. Larsson, F. Harrysson, B. Olsson, and J. Berg. “An outdoor-to-indoor prop-

agation scenario at 28 GHz ’‘. In 8th European Conference on Antennas and Prop-

agation (EuCAP), April, 2014.

[244] H. Zhao, R. Mayzus, S. Sun, M. Samimi, J. K. Schulz, Y. Azar, K. Wang, G. N.

Wong, F. Gutierrez, and T. S. Rappaport. “28 GHz millimeter wave cellu-

lar communication measurements for reflection and penetration loss in and

around buildings in New York city ”. In IEEE International Conference on Com-

munications (ICC), June, 2013.

[245] F. S. Chaves and K. Bechta. “5G network deployment: Interplay of key ele-

ments in the challenging outdoor-to-indoor scenario,”. In European Conference

on Networks and Communications (EuCNC), June, 2016.

[246] W. Karner, A. Paier, and M. Rupp. “Indoor coverage prediction and optimiza-

tion for UMTS macro cells,”. In 3rd International Symposium on Wireless Commu-

nication Systems (ISWCS), September, 2006.

[247] C. Seltzer. “Indoor coverage requirements and solutions,”. In IEE Colloquium

on Antennas and Propagation for Future Mobile Communications, pages 3/1–3/4,

London, February February, 1998. IET.

[248] D. Hong, S. Choi, and J. Cho. “Coverage and capacity analysis for the multi-

layer CDMA macro/indoor-picocells”. In IEEE International Conference on Com-

munications (ICC), volume 1, pages 354–358, June, 1999.

122

References

[249] H. Andersson, R.S. Karlsson, P. Larsson, and P. Wikstrom. “Improving system

performance in a WCDMA FDD network using indoor pico base stations”. In

IEEE 56th Proceedings Vehicular Technology Conference, pages 467–471, Septem-

ber, 2002.

[250] K. Hiltunen, B. Olin, and M. Lundevall. “Using dedicated in-building systems

to improve HSDPA indoor coverage and capacity,”. In 61st IEEE Vehicular Tech-

nology Conference, volume 4, pages 2379–2383. IEEE, June June, 2005.

[251] Z. Uykan and K. Hugl. “HSDPA system performance of optical fiber dis-

tributed antenna systems in an office environment,”. In IEEE 16th International

Symposium on Personal, Indoor and Mobile Radio Communications, volume 4, pages

2376–2380, Berlin, September, 2005.

[252] J. Borkowski, J. Niemela, T. Isotalo, P. Lahdekorpi, and J. Lempiainen. “Utiliza-

tion of an Indoor DAS for Repeater Deployment in WCDMA,”. In IEEE 63rd

Vehicular Technology Conference, pages 1112–1116, Melbourne, May May, 2006.

IEEE.

[253] D. J. R. Martin. “Leaky-feeder radio communication: A historical review,”. In

34th IEEE Vehicular Technology Conference, volume 34, pages 25–30. IEEE, May

May, 1984.

[254] K. J. Bye. “Leaky-feeders for cordless communication in the office,”. In 8th Euro-

pean Conference on Electrotechnics. Conference Proceedings on Area Communication,

pages 387–390, Stockholm, June, 1988.

[255] E. Mutafungwa, Z. Zheng, J. Hämäläinen, M. Husso, and T. Korhonen. “Ex-

ploiting femtocellular networks for emergency telemedicine applications in in-

door environments,”. In 12th IEEE International Conference on e-Health Network-

ing Applications and Services (Healthcom), pages 283–289, July, 2010.

[256] E. Mutafungwa, Z. Zheng, J. Hämäläinen, M. Husso, and T. Korhonen. “On

the use of Home Node Bs for Emergency Telemedicine Applications in Various

Indoor Environments,”. International Journal on E-Health and Medical Communi-

cations, 2:91–109, March, 2011.

[257] P. R. Devadoss and S. L. Pan. “Leveraging eGovernment Infastructure for Crisis

Management: Lessons from Managing SARS Outbreak in Singapore,”. Journal

of Information Technology Theory and Application, 6, 2004.

[258] Y. Xue and H. Liang. “IS-Driven process engineering: China’s public health

emergency response to SARS crisis,”. Journal of Information Technology Theory

and Application, 6, 2004.

[259] B. V. D. Walle and M. Turroff. “Emergency Response information sys-

tems: emerging trends and technologies, ”. Communications of the ACM, 50,

March, 2007.

[260] T. Gomes et al. “A survey of strategies for communication networks to protect

against large-scale natural disasters,”. In 8th International Workshop on Resilient

Networks Design and Modeling (RNDM),, September, 2016.

[261] A.J.J. Sluyter. “The role of Communication Systems in Emergency Medical

Services”. IEEE Transactions on Vehicular Technology, 25:175–186, 1976.

123

References

[262] D. Ziadlou, A. Eslami, and H. R. Hassani. “Telecommunication methods for

implementation of telemedicine systems in crisis”. In Third International Con-

ference on Broadband Communications, Information Technology & Biomedical Appli-

cations, pages 268–273, November, 2008.

[263] C.S. Pattichis, E. Kyriacou, S. Voskarides, M.S. Pattichis, R. Istepanian, and C.N.

Schizas. “Wireless Telemedicine Systems An Overview”. IEEE Antennas and

Propagation Magazine, 44:143–153, August, 2002.

[264] W. Y. Lin, Y. C. Chen, R. Y. Chang, S. H. Chen, and C. L. Lee. “Rapid WiMAX

network deployment for emergency services,”. In 7th International Symposium

on Wireless and Pervasive Computing (ISWPC), July, 2012.

[265] S. Allsopp. “Emergency airborne 4G comms to aid disaster traffic manage-

ment,”. In Road Transport Information and Control Conference (RTIC), Octo-

ber, 2014.

[266] L. G. Askoxylakis, A. Makrogiannakis, A. Miaoudakis, S. Papadakis, N. E.

Petroulakis, M. Surligas, A. Traganitis, and N. Vervelakis. “A Rapid Emer-

gency Deployment mobile communication node,”. In IEEE 19th International

Workshop on Computer Aided Modeling and Design of Communication Links and

Networks (CAMAD), December, 2014.

[267] T. Sakano. “Bringing movable and deployable networks to disaster areas: de-

velopment and field test of MDRU,”. IEEE Network, 30:86–91, January, 2016.

[268] D. Iland and E. M. Belding. “Emergenet: robust, rapidly deployable cellular

networks,”. IEEE Communications Magazine, 52:74–80, December, 2014.

[269] Q. T. Minh, K. Nguyen, C. Borcea, and S. Yamada. “On-the-fly establishment

of multihop wireless access networks for disaster recovery,”. IEEE Communica-

tions Magazine, October, 2014.

[270] X. Chu et al. Heterogeneous Cellular Networks: Theory, Simulation and Deployment.

Cambridge University Press, United Kingdom, 2013.

[271] Altair, Winprop overview, http://www.altairhyperworks.com/product

/FEKO/WinProp.

[272] “LTE; Evolved Universal Terrestrial Radio Access (E-UTRA); Radio Frequency

(RF) system scenarios (Release 9), ” 3GPP Tech. Rep. 36.942 v9.0.1, April. 2010.

124

-otl

aA

DD

5

21/

810

2

+gggai

a*GMFTSH

9 NBSI 6-6608-06-259-879 )detnirp( NBSI 3-7608-06-259-879 )fdp( NSSI 4394-9971 )detnirp( NSSI 2494-9971 )fdp(

ytisrevinU otlaA

gnireenignE lacirtcelE fo loohcS gnikrowteN dna snoitacinummoC fo tnemtrapeD

if.otlaa.www

+ SSENISUB YMONOCE

+ TRA

+ NGISED ERUTCETIHCRA

+ ECNEICS

YGOLONHCET

REVOSSORC

LAROTCOD SNOITATRESSID

hall

U ma

nIdn

a ed

oce

D dn

abnI

rof

sdoh

teM

noit

agit

iM

ecne

refr

etnI

dna

noi

taco

llA

ecru

ose

R fo

sisy

lan

A ec

namr

ofre

P

gni

yale

R dr

awr

oF y

tisr

evi

nU

otla

A

8102

gnikrowteN dna snoitacinummoC fo tnemtrapeD

fo sisylanA ecnamrofreP dna noitacollA ecruoseR

noitagitiM ecnerefretnI dnabnI rof sdohteM drawroF dna edoceD

gniyaleR

hallU manI

LAROTCOD SNOITATRESSID

-otl

aA

DD

5

21/

810

2

+gggai

a*GMFTSH

9 NBSI 6-6608-06-259-879 )detnirp( NBSI 3-7608-06-259-879 )fdp( NSSI 4394-9971 )detnirp( NSSI 2494-9971 )fdp(

ytisrevinU otlaA

gnireenignE lacirtcelE fo loohcS gnikrowteN dna snoitacinummoC fo tnemtrapeD

if.otlaa.www

+ SSENISUB YMONOCE

+ TRA

+ NGISED ERUTCETIHCRA

+ ECNEICS

YGOLONHCET

REVOSSORC

LAROTCOD SNOITATRESSID

hall

U ma

nIdn

a ed

oce

D dn

abnI

rof

sdoh

teM

noit

agit

iM

ecne

refr

etnI

dna

noi

taco

llA

ecru

ose

R fo

sisy

lan

A ec

namr

ofre

P

gni

yale

R dr

awr

oF y

tisr

evi

nU

otla

A

8102

gnikrowteN dna snoitacinummoC fo tnemtrapeD

fo sisylanA ecnamrofreP dna noitacollA ecruoseR

noitagitiM ecnerefretnI dnabnI rof sdohteM drawroF dna edoceD

gniyaleR

hallU manI

LAROTCOD SNOITATRESSID

-otl

aA

DD

5

21/

810

2

+gggai

a*GMFTSH

9 NBSI 6-6608-06-259-879 )detnirp( NBSI 3-7608-06-259-879 )fdp( NSSI 4394-9971 )detnirp( NSSI 2494-9971 )fdp(

ytisrevinU otlaA

gnireenignE lacirtcelE fo loohcS gnikrowteN dna snoitacinummoC fo tnemtrapeD

if.otlaa.www

+ SSENISUB YMONOCE

+ TRA

+ NGISED ERUTCETIHCRA

+ ECNEICS

YGOLONHCET

REVOSSORC

LAROTCOD SNOITATRESSID

hall

U ma

nIdn

a ed

oce

D dn

abnI

rof

sdoh

teM

noit

agit

iM

ecne

refr

etnI

dna

noi

taco

llA

ecru

ose

R fo

sisy

lan

A ec

namr

ofre

P

gni

yale

R dr

awr

oF y

tisr

evi

nU

otla

A

8102

gnikrowteN dna snoitacinummoC fo tnemtrapeD

fo sisylanA ecnamrofreP dna noitacollA ecruoseR

noitagitiM ecnerefretnI dnabnI rof sdohteM drawroF dna edoceD

gniyaleR

hallU manI

LAROTCOD SNOITATRESSID